Studies in Variation, Contacts and Change in English
Looking for rhetorical thresholds: Pronoun frequencies in political speeches
The deliberate and considered use of personal pronouns is one of the primary linguistic features used by political speakers to manage their audiences’ perceptions of in-groups and out-groups. In this diachronic study of political speeches over the last two centuries, I will argue that a notable shift took place in politicians’ use of personal pronouns around the 1920s, immediately following the time broadcast media emerged on the scene.
Walter Lippmann (1889–1974) was an American journalist, author and political thinker. In Public Opinion, his widely influential analysis of contemporary politics and the state of democracy, Lippmann argued that most people are not capable of forming a comprehensive and coherent view of the increasingly complex world around them, and that consequently they need need to have their opinions and views guided and managed by the well-informed. Crucially, Lippmann suggested that these well-informed individuals are usually not the leaders, and that the processes of analysis and political leadership are two separate skills. What civil society needs, Lippmann argued, are analytical insights into the issues at hand followed by political rhetoric that sells the resulting new policies to the public. Consequently, the creating and managing of consent by means of carefully crafted process of analysis and mass communication had come to a crossroads in the early twentieth century, as the emerging new technologies and more studied use of mass psychology were about to change the very nature of democracy — and political language.
In this paper, I will argue that the early twentieth century witnessed a stylistic change in how politicians addressed the public, and that this change was at least partly, if not predominantly driven by the two mechanisms that Lippmann identified: modern means of communication and more sophisticated psychological insights into the creation of public consent. As the means of reaching ever-wider audiences improved, the language of politics had to shift to be more inclusive, at least on the surface, and to pay more attention to how diverse audiences might react to the message. From the beginning of the nineteenth century to the end of the twentieth century, Anglo-American societies witnessed extensive political, societal, technological and cultural changes which could be collectively described as the process of democratization. As the electorate expanded and modern mass media developed, politicians wishing to win over the voting public had to adjust their message. Although political persuasion is naturally highly contextual, there are also rhetorical means which appear to remain relatively stable in function, if not in frequency.
The use of personal pronouns has been shown in numerous previous studies of political language to be the single most important linguistic feature when it comes to fostering and managing sentiments of group membership (see section 3). Given that nearly all political speeches are ultimately intended to exploit in-group and out-group dynamics to the advantage of the speaker and the detriment of their political opponents, the considered use of personal pronouns is a highly salient characteristic of political speaking. However, to my knowledge no previous study has looked at the phenomenon from a frequency-based, diachronic perspective in an effort to discover whether the patterns and practices of personal pronoun use in political speeches constitute a stable and reasonably unchanging phenomenon, or whether there are discernible trends that transcend the styles of individual speakers and the demands of specific issues.
In this pilot study I will combine methods of diachronic corpus linguistics and critical discourse analysis to examine how often and in what part of the speech politicians tend to refer to us, them and the other, and whether we can discern diachronic trends which might be related to the process of democratization.  As a point of departure to most studies of political language, which have traditionally focused on specific case studies ranging from well-defined political scenarios to individual politicians and sometimes single influential speeches, my starting point will be somewhat different.Using the Small Corpus of Political Speeches as my primary data (see section 4), I will examine speeches as a developing text type over the time period of 1800–2000 from a diachronic perspective without focusing on, and most of the time without even acknowledging, the topics of the individual speeches. Although the corpus does include metadata on the speakers, their party affiliations, the topics of the individual speeches, and on various other factors, these will be set aside with the exception of specific examples, which will be contextualized. I will present baseline evidence of the overall use of personal pronouns in political speeches and show that while the overall personal pronoun frequency has remained largely stable over the past two centuries, political rhetoric has in fact undergone a significant change, culminating in a rhetorical threshold in the early twentieth century when political speakers switched from an I-focused mode of speech to a we-focused one, and the compound frequency of references to self and references to the in-group increased markedly at the expense of other personal pronouns.
Rhetoric, the art of public speaking, is one of the oldest and most influential areas of intellectual engagement. From Aristotle and Cicero to St. Thomas Aquinas to Dryden, the formal study of rhetoric has been a key topic of study in the humanities. Although the practice of rhetoric has undergone gradual changes over the centuries, most notably when it comes to changes in the media of communication and the consequent increase in the size of the potential audience, the fact remains that an effective political speaker, or their team of writers, needs to be able to craft and deliver a convincing and compelling public speech.  The success of a political speech does not hang on the content of the message alone, but rather relies on the delivery, the framing of concepts and the associations and affiliations that the speaker manages to evoke. Like the dramatic soliloquy, a political speech is an archetypal example of a text written-to-be-spoken. With few exceptions, political speeches are not ad hoc presentations, regardless of how they come across — or, as Mark Twain is said to have quipped, “it usually takes more than three weeks to prepare a good impromptu speech”. Political speeches are prepared in advance, today often by teams of professional speechwriters, they are tested using focus groups, and they follow tried-and-tested rhetorical techniques. In short, political speeches are highly engineered texts designed to affect the listener and to elicit a specific emotional and intellectual response, which later translates into a vote. The speeches studied in this paper were primarily delivered by prominent politicians such as party leaders, presidents, presidential candidates, and government ministers, with a smattering of civil rights leaders and minor members of parliament thrown in for good measure. Consequent to the high status of most of the speakers, it is likely that most of the speeches were the product of team effort.
In the present study, the objective of the research project is less critically oriented than in most studies of political language. Linguistic scholarship of political discourse has largely been conducted within the framework of Critical Discourse Analysis, wherein the established methodology has focused on close reading individual speeches to uncover veiled meanings, doubletalk, metaphorical tropes, stock phrases, semantic prosodies and other linguistic mechanisms by which political actors influence each other and their audiences, and by which power is used and abused in society (see, e.g., Fairclough 1989, 2001).  Although I will comment on individual speeches on occasion, the main focus will be on the long-term trends we can observe in the use of personal pronouns in political speeches over a period of some two hundred years. 
Of all human activities, the field of politics is perhaps the most susceptible to the phenomenon of mediatization, that is, of the processes by which media has increasingly become the dominant force in society and as a result of which communicators must take media and their inherent rules, the so-called media logic, into consideration when formulating their message.  This not only means that a political speaker would prepare their message to fit the audience in front of them, but also that the core dynamic of political communication may change away from the message and toward the delivery. Observant of the “logic of the media”, a concept coined by Hjarvard (2008), politicians and other actors in society, such as entertainment celebrities and religious figures, adopt new objectives and discard ones that no longer serve their purpose. According to Esser and Strömbäck (2014: 7), “As politics becomes increasingly mediated, it becomes more important for political actors and institutions to use the media to reach out to larger groups of society”. In the present study, I will specifically consider the impact that the transition from paper-based mass media to radio and television and finally to Internet media had on pronominal references in political speaking, proceeding from the hypothesis that the enlargement of the audience would change the relationship between the speaker and the audience by focusing more attention on the need to foster a sense of community rather than presenting the politician as a charismatic authority.
The primary objective of political speaking is to affect the opinions and choices of others. This usually involves the juxtaposing of opposing views and, consequently, the construction or maintenance of listeners’ sense of having affiliations and allegiances with a particular community (Hahn 2003, Allen 2007). Although the concept of community can be difficult to define in precise terms, one of the key defining characteristics is that the members of a community share a sense of belonging to that group and not to another (Íñigo-Mora 2004: 29). Consequently, one of the primary techniques in political rhetoric is the setting up of contrasts between us and them, or between the in-group and the out-group. Depending on the political arena, the latter may be portrayed not merely as the Other but as a Threat or an Enemy (see, e.g., Pennycook 1994, van Dijk 1997a, Joseph 2006).
As Chilton (2003) notes, a variety of linguistic devices ranging from the deliberate choice of register-specific words to the use of address terms can be employed for the purposes of signalling distance or affiliation between political actors and for establishing and maintaining individual and group identities. The effective use of indexicality, defined by Chilton as “the implicit signalling of political affiliations”, is one of the central features of political language use, and pronouns have been recognised by scholars as one of the most powerful means of enacting indexicality (see, e.g., Brown and Gilman 1960, Maitland and Wilson 1987, Pennycook 1994, Allen 2007, Petersoo 2007, Fetzer and Bull 2008, Cramer 2010, Zhan 2012, Bello 2013, Chimbarange et al. 2013 and Roitman 2014). 
Paradoxically, the rhetorical usefulness of personal pronouns derives from their inherently affective qualities as well as from their referential ambiguity, and thus two major approaches to the study of pronouns in political language can be distinguished.  The first concerns the use of pronouns for the purposes of distinguishing in-groups and out-groups, and the other the use of pronouns for deliberately obfuscating or misdirecting audiences’ allegiances. The use of pronouns for fostering a sense of belonging to an in-group or, conversely, for the purposes of othering has been the subject of numerous studies over the years, and the general principles by which such mechanisms function can be considered established in literature. Although the “purposeful use of pronouns”, as described by Thomas and Wareing (1999), is a readily identified feature of all public communication and not just of political speaking, it may be argued that personal pronoun use is never more intensively purposeful than in political rhetoric. Consider the following example from Sarah Palin’s speech at the Republican Governors Association in 2008 during her brief governorship of Alaska:
In the first sentence of this passage, Palin makes a prediction, framed as a promise, to the audience about what will transpire in the near future. The sentence begins with a strong declarative speech act “I promise you” which fosters a positive relationship between the speaker and the audience. Next, the reference to “their governors” posits the people, “Americans”, as the Other in relation to the current audience, Republican governors and their staff, and “our abilities” at the end of the sentence reinforces the in-group mentality. The next sentence creates a new Other, “the pundits”, casting them in a negative light (“their idle talk”). The next to sentences use “our” and “we” in reference to the positive qualities and bright future of the in-group, while the last sentence creates a third out-group, “the new Congress and President”, suggesting that the in-group of governors is superior to the new out group. Thus, within five sentences, Palin uses their and them three times, each time in reference to a different group of outsiders.
The second strand of research focuses on the use of pronouns as devices of deception by deliberate and sometimes complicated shifting of referents within the single discourse event. Fetzer and Bull (2008: 275) note that although pronouns are linguistically unambiguous in the sense that their referent can always be retrieved and the noun phrase placed in their stead, the same cannot be said when it comes to the discursive perspective. According to Allen (2007: 12), “Shifting identity through pronoun choice and using pronouns with ambiguous referents enables politicians to appeal to diverse audiences which helps broaden their ability to persuade the audience to their point of view. It is a scattergun effect – shoot broadly enough and you’ll hit something”.
The prime example of this inherent ambiguity of personal pronouns can be seen in the first person plural we, which Pyykkö (2002: 246) describes as “a central political force of influence” and Wales (1996: 58) defines as “usefully ambivalent” for political speaking. It is of course well attested that the English we can be used both inclusively and exclusively, as a vague in-group reference to hide behind, and sometimes with multiple referents within a single sentence. Quirk et al. (1985: 350–1) distinguish between no less than eight senses of we: the generic, the inclusive authorial, the editorial, the rhetorical, in reference to the hearer, in reference to a third party, the royal we, and the self-referential we. Out of these, it is the rhetorical we that is so strongly associated with political language use. According to Wales (1996: 59), “the politician-speaker often uses we with the double inference and presumption that he or she is not only speaking on behalf of the party or government (exclusive), but also on behalf of the audience (inclusive)”. To illustrate this with an example drawn from the corpus to be discussed shortly, let us take an example. In the following speech delivered by President Obama in Denver, 9th of July, 2014, we see a shift in person focus in nearly every sentence, sometimes multiple times within a single sentence.
The speech is not confusing or incoherent in any way, but the repeated switches from I to we to you to my and back again build up to an intense sense of linking the speaker (the president) with the audience and thereby creating what Lippmann described as a “community of feeling”.
However, sometimes the shifts in focus can indeed be confusing. In the following excerpt from Donald Trump’s speech delivered on 30th December, 2015, in Hilton Head, South Carolina, there are two different, juxtaposed referents to the pronoun they: the media and Trump’s own supporters.
Having thus established that politicians make extensive and studied use of personal pronouns, it is time to turn to the question at hand, namely can we discern any diachronic changes in the political use of personal pronouns?
The present study takes a corpus linguistic and data-driven approach to the issue of pronoun use in political speaking. Consequently, the emphasis of the analysis relies much more on frequency-based observations of occurrence and distribution than what is typical to political discourse analysis. Here and there, representative examples will be used to illustrate particular linguistic practices, but the extracts were not selected to exemplify specific political arguments nor are they contextualized in great detail.
I use as primary data the Small Corpus of Political Speeches (SCPS), compiled over several years at the Universities of Helsinki and Tampere. The corpus was originally conceived of as a teaching tool for corpus linguistic methodology classes. During each course, the students were presented with the corpus, its compilation principles and current composition. The class would then discuss how to expand the corpus — for example, one course decided to add more speeches by female politicians while another wanted to make it more representative of nineteenth-century American speeches. The students would then each download a set number of speeches available from various online speech archives, edit the speeches according to a set of principles. The following week, the new and enlarged version of the corpus would then be used in a small research project. In 2013 and 2014, the corpus was enlarged further with the assistance of two research assistants, Veera Saarimäki and Jenni Riihimäki respectively, at which point new, more carefully researcher metadata was also added.
In the present study, I used version 4 of the corpus, comprising 875 speeches and a total word count of 3.8 million words. The corpus covers the period 1800 to 2010 with a slight bias towards the turn of the twenty-first century, and focuses on American and British speeches but also includes a selection of Australian speeches. A wide range of topics and speech situations are included with no apparent bias for any specific type, though due to sourcing issues there is undoubtedly some bias towards well-known politicians. Similarly, although care was taken to avoid diachronic selection bias, the further back we go the more bias there is in favour of prominent politicians and historically notable speeches, such as presidential inaugural addresses. All the speeches are included in full length. The mean length is 3,880 words and the median length 3,324 (see Figure 1). The one outstanding outlier is Ted Cruz’s filibuster speech against Obamacare, deliver on September 24, 2013, which runs to 76,360 words. There is no significant change over the timeline when it comes to speech length (r=-0.06, R2=0.004). For more on the corpus and its metadata, see its entry in the Corpus Research Database.
As noted above, the speeches in the corpus were obtained from a variety of online archives of political speeches.  Consequently, the speeches have been transcribed by a wide variety of transcribers following an equally diverse variety of conventions; some recorded audience responses (applause, laugher, etc.) as well as meta-discourse by the speaker (such as repeated expression of “thank you” as a response to prolonged applause), while others only include the actual speech. To the extent possible, only the actual speech is included in the corpus and all other text in the source file is included within tags, which makes it easy to ignore them in queries. Likewise, the original transcription conventions differ when it comes to accurately features of spoken language are rendered in writing. Some transcribers expand features such as contracted forms of auxiliary verbs, while others treat them as a single lexical item: e.g, have not vs. haven’t.  Although this would not be an issue for a human reader, it makes a notable difference in corpus linguistic analysis when it comes to word counts, retrieval strings, collocate analysis, etc. Thus, in the interest of the present analysis, all contracted forms on the stop word list were automatically expanded, e.g., can’t -> cannot, haven’t -> have not, etc.
When it comes to closed-class lexical items capable of signalling the presence of one or more human actors in the discursive context, the English pronoun system provides us with personal pronouns proper, both subject and object, possessive pronouns, and reflexive pronouns. Given that in the present study we are primarily interested in references by the speaker either to himself or herself, or to others, it makes sense also to include possessive determiners associated with personal pronouns because they are high frequency items and they can be used to signal political associations similarly to personal pronouns. On the other hand, both possessive pronouns and reflexive pronouns are low frequency items and virtually never used without the corresponding personal pronoun in close proximity, and consequently they were left out of the study.
Given that the corpus is restricted to the single register of political speeches, it seems reasonable to make certain blanket assumptions about the functional aspects of the personal pronouns. For example, although it is possible that some politicians in the data might used the royal or authorial we in reference to themselves, for the sake of keeping the analysis manageable I will make the assumption that such usage is rare enough to be negligible. Likewise, while the use of the singular gender-neutral they has become more common in recent years, it must be considered extremely rare over much of the timeline.
The analysis was operationalised as follows. The corpus was first queried for all cases of personal pronouns (both subject and object), possessive determiners and possessive pronouns Next, the pronouns and determiners were classified into semantic groups according to their primary referent, considering the first person singular to be Self-referential, the second person singular and plural as Audience-referential, the first person plural as Inclusive-referential and the third person singular and plural as Other-referential (Table 1). As the table shows, Audience references are the least frequent type of person pronominal use at a mean frequency of 5.76/1,000 words, while Inclusive references are the most common at 22.59/1,000 words.
However, as can be seen in Table 2, some speakers will far exceed these baseline frequencies, at least on occasion. Interestingly perhaps, many of the names among the top ranks are highly recognisable politicians, which suggests that the high frequency of personal pronouns may be a part of a successful rhetorical strategy, or at the very least it is not a hindrance.
Importantly, it goes without saying that these referential categories are only a rough approximation made necessary by the relatively large-scale of the corpus-based analysis. For example, it is certainly possible to find individual instances where references to him or them could be considered Inclusive, such as when a politician is thanking a specific supporter of his or her campaign, and likewise there may well be occasions where references to you may address a foreign leader who is not present and who may therefore be understood as the Other by both the speaker and the audience. However, while it would be possible to carry out a more detailed analysis where every instance of a personal pronouns was individually categorized into one of the semantic groups, doing so would involve analysing hundreds of thousands of items, which was not possible in the scope of the present study. Nevertheless, I would argue that the overall impression derived from this rough categorization is valid for identifying trends and general tendencies in political language.
To begin with a general statement, the overall frequency of personal pronouns shows a slight positive cline over the timeline (Figure 2).  If treated as a linear correlation, the trend is not particularly strong (Pearson’s r=0.21, R2=0.04), but it does suggest that the politicians of our time tend to use personal pronouns somewhat more frequently than the politicians of earlier eras, especially going back a century or more.
However, when we start focusing on specific semantic categories of pronouns, things get much more interesting. For clarity, each of the following figures shows two categories of pronouns only, starting with self references and audience references (Figure 3). The polynomial lines are fitted to show trends more clearly.  In the scatter plots, the dots represent individual speeches, with the horizontal axis denoting time and the vertical axis denoting standardized frequency per thousand words.
The overall pattern shows that references to self and to the audience are both at their most frequent during the latter half of the nineteenth century, going into a slight decline at the turn of the century. The change in the frequency of audience references is slight, however. By contrast, the decline in self-references during the early part of the twentieth century was likely motivated by broad changes in democratic society, where self-promotion was increasingly disliked and instead politicians had a greater need to present themselves as representatives and members of the electorate.
This may also be interpreted at least in part as a reflection of the changing media landscape. As political speeches were increasingly broadcast to non-immediate audiences, that is, to audiences who may be listening to the speech by radio or watching it on television, the lack of proximity somewhat dilutes the effect of emotive and affective self-reference. The newfound incline in self-references observed from the 1970s onward may in turn signal a change to a more person-centred style of politics and, though this is more hypothetical, improvements in audiovisual technology which allow close-up images of speakers to be broadcast, making the communicative style more immediate and thus making the use of self-references more effective. 
Clear evidence of the shift from person-centred rhetoric to a more group-centred style is seen in Figure 4. The frequency of inclusive references using the group indexical pronouns we, us and our increases dramatically at the same time as self-references go into decline. Notably, even though self-references turn back into an incline toward the end of the twentieth century, inclusive references continue to increase even further at the same time.
The frequency change in inclusive references is the most dramatic difference observed in the use of personal pronouns. As the scatterplot shows (Figure 4), this does not mean that there were no speakers who used inclusive references in high frequencies during the nineteenth century, but it is clear that doing so was much less common then than it is today. Let us consider two somewhat extreme examples, Thomas Babington Macaulay’s 1852 address in Edinburgh on his re-election to parliament, and Gordon Brown’s 2004 speech at the Labour party conference in Brighton as Chancellor of the Exchequer.
Macaulay’s speech is clearly highly personal in orientation. The speaker’s personal opinions and views are communicated directly — “And I will tell you why I entertain this opinion”, “I know him well by character” — and the speech maintains a single point-of-view, namely the speaker’s, and relies entirely on the speaker’s personal prestige and charisma.
Brown’s speech, by contrast, makes use of what Fetzer and Bull (2008: 281) describe as “the communicative strategy of over-inclusion”. By switching between his own thoughts and those of the party, the speaker first creates unity between himself, the party and the country, conceptually presenting all three as aspect of the same: “But for me, for us the Labour party, and for the country, this is not enough”. As Moberg and Eriksson (2013: 320) argue, “By selecting we rather than another pronominal form a speaker introduces a bond with her interlocutors. Through this other persons are brought into an obligation pattern and the speaker speaks as a representative of a group rather than as a self.” Once this is done, Brown continues with a series of boasts about the successes of the trinity, each followed by an aspirational statement promising that even more will be done in the future.
Although the two excerpts above already exemplify the difference between nineteenth and twentieth century speaking styles, there is one specific category of speech where the diachronic stylistic switch is particularly salient, namely the inaugural address of American presidents. Figure 5 shows the dramatic change: while nearly all early inaugural addresses were focused on the president himself, after Benjamin Harrison in 1889, there is only one inaugural speech, Herbert Hoover’s in 1929, in which the newly elected president used self-references more than inclusive references to the country.   Hoover uses the first person singular 53 times and the first person plural only 11 times.
The overwhelmingly clear stylistic shift that takes appears to begin at the end of the nineteenth century onward can be explained both by the processes of democratization and mediatisation. Following the Civil War in 1861 to 1865, there was considerable need to bring the country together and to use a more inclusive voice. While the first presidents frequently used the inaugural address to speak about themselves and the responsibilities of the office, presidents of the post-Civil War era started focusing more on the country.  The following quote from the beginning of Thomas Jefferson’s inaugural address in 1801 demonstrates the personal and modest style of the early speeches.
By contrast, the next example from the beginning of Bill Clinton’s first inaugural address shows the new inclusive style in full effect.
This need for inclusivity was intensified further during the World Wars, after which time it appears to have become a fixed standard. Significantly, the same time period also witnessed the appearance of broadcast media in the form of radio and television, and, in more recent times, of online media.
Returning to the full dataset, the plot for Inclusive and Audience references is given in Figure 6. Both types of reference can be used for essentially the same purpose, addressing the audience directly, but while the former includes the speaker as a member of the same in-group, the latter posits a dialogic relationship between the speaker and the audience, as seen earlier in Figure 3. Looking at Figure 6, we see that the frequency of Audience references has remained quite stable over the two-hundred-year timeline, while Inclusive references have increased, as noted earlier.
In the following example from Donald Trump’s 2015 speech announcing his candidacy for President, Trump switches repeatedly between references to Self, the Audience and two in-groups: the Trump organisation and the people. This establishes a subtle message, which links the two in-groups together, suggesting that by siding with Trump, the audience can share the success of the Trump business empire. The direct Audience references come toward the end of the excerpt, when Trump starts addressing the audience directly as he starts talking about the sorry state of infrastructure in America. By using the second person plural (“you have all these disastrous airports”), he emphasises the difference between the Audience’s low-quality airports and the luxurious airports of the foreigners. Importantly, while Trump says “we’re in a third world country” twice, fostering a sense of in-group camaraderie, he never once uses we or our when talking about American airports because that could intuitively suggest that Trump, too, has conceptual ownership of them.
The final plot (Figure 7) shows the relationship between inclusive references and references to the Other (they, them, their, he, his, him, she, her). Here, again, we see a shift in the early 1900s, when references to the in-group overtake references to those outside it. Note the consistency of the difference, as attested by the scatter plot: nearly all nineteenth century speeches feature a higher frequency of Other references and nearly all speeches after 1950 feature a higher frequency of in-group references.
The following excerpt from Anna Howard Shaw’s 1915 speech, delivered in New York City at the New York State equal suffrage campaign, serves as a good example of very frequent use of they:
As the previous figures illustrate, self-references follow a pattern of decline from the beginning of the nineteenth century all the way to approximately 1960s, where they appear to turn into an incline. Inclusive references, by stark contrast, experience an incline during the same period, reaching a plateau in the 1960s. The two trend lines cross at the turn of the century, but it is safe to say that the main shift really takes place in the first twenty years of the twentieth century. Audience references remain almost unchanged throughout the two centuries.
In addition to the needs of addressing the immediate audience in a compelling fashion, the rise of Inclusive references in particular can also be linked to the emergence of national and international political entities as key players in the political field during the twentieth century. In such contexts, the conceptual in-group can comprise nations states, political or institutional entities, which are pronominally referred to in speeches as if they were animate entities. To exemplify this in action, let us finish with an excerpt from Margaret Thatcher’s 1979 speech on foreign policy, the newly appointed Prime Minister outlines international relations by strongly aligning Britain with the United States and Europe.
Thatcher uses the first person plural in references to the in-group, or “the West”, while reserving the second person plural for the out-group, “the Third World”. At the end of the excerpt, Thatcher repeatedly alternates between we and them as she argues that the West “supplies” and “provides” the Third World with everything.
This exploratory study takes a first step toward a more comprehensive description of quantitative changes in Anglophone political speeches during the nineteenth and twentieth centuries.  The results show that the use of personal pronouns and possessive determiners has remained relatively unchanged over the timeline with one significant exception, namely, that the frequency of the inclusive references increases dramatically from the early twentieth century onward, coinciding with the introduction and rapid expansion of electronic mass media (Figure 8).
The strength of the phenomenon is such that it merits the label rhetorical threshold. The evidence from SCPS suggests that the beginning of radio broadcasting coincides, if not ushers in, the beginning of a more we-focused style of political speaking, which occurs concurrently with a temporary decline of speaker-focus. Following the Second World War and the introduction of television, self-references by political speakers turn back into an include, while inclusive references also continue to rise. The data is too scarce to allow more than anecdotal conclusions about the era of digital media, but there is some indication that the self-referential style may be turning into a decline again. Notably, there is also weak evidence to suggest that audience references may have been on a slight increase since the beginning of the Internet era.
Although this empirical finding can be interpreted as indirectly supporting early twentieth-century observations about a contemporaneous transition from an individuated to a communal sense of self as the predominant mode of personal identity around the turn of the century, it is important to note that there are limits to the extent to which can read changes in political language as reflecting and taking advantage of the contemporary zeitgeist.  Indeed, the dramatic increase in the use of we can also be interpreted as supporting the conclusion of earlier research into pronoun use, in which it has been argued that in carefully crafted political discourse, inclusive references are deliberately used to misdirect the listener and to leave the burden of interpretation to the audience.
 A more specialized approach to the topic is Political Discourse Analysis, defined by van Dijk (1997a) as a field that combined an essentially impartial linguistic analysis of political discourse with critical elements associated with Critical Discourse Analysis. [Go back up]
 It is important to note that the present study focuses exclusively on speeches delivered by politicians and political activists in political contexts. The study does not include political interviews, biographies and autobiographies, articles, columns, blogs or tweets. [Go back up]
 For a general introduction, see Esser and Strömbäck (eds.) (2014); for an excellent introduction to mediatization in the linguistic framework, see Androtsopoulos (2014); and for an introduction to media influence on language use, see Sayers (2014). [Go back up]
 Although the rest of the discussion will concern primarily personal pronouns, it is worth noting that pronouns other than personal have been studied from the CDA and sociolinguistic perspectives. For example, Acton and Potts (2014) shows how Sarah Palin’s use of demonstratives helped foster a sense of shared experience with her audience. [Go back up]
 The sources are given in the corpus metadata which is always distributed with the corpus. All the online archives used are available open access. [Go back up]
 The older speeches typically follow a more formal convention. In many cases this reflects the fact that the speech was not transcribed live but instead distributed in writing after the event. [Go back up]
 Note that the fitted lines in all the scatter plots are polynomial. [Go back up]
 When it comes to diachronic linguistic data, nonlinear fitted lines are nearly always preferable to linear fits. As clearly demonstrated by the graphs used in this paper, linear fits would grossly misrepresent periodic and oscillating phenomena. [Go back up]
 Fox and Park (2006), studying pronoun choice among embedded and nonembedded war reporters during the 2003 Iraq War, show that reporters appeared to communicate their objectivity by careful balancing I and we references. Likewise political speakers must walk a fine line between appearing to share their own thoughts and thereby appearing intimate and genuinely human, and framing the discourse as a joint cause by evoking the inclusive group indexical we. As the evidence shows, few politicians do only one, most opting to use both strategies. [Go back up]
 Watch video footage of President Herbert Hoover from 1929 [Go back up]
 Hoover’s inaugural address was the first to be recorded on a newsreel. [Go back up]
 For valuable discussion of this shift, see Kovalyova (2005). Regarding the style of the early presidents, Kovalyova notes that ”regardless of the number of points an [Inaugural Address] covered, the figure responsible for their implementation remained the president himself. That is why entering the highest office of the land, he talked – first and foremost – about his duties” (41). [Go back up]
 These issues will be examined in the Democratization, Mediatization and Language Practices in Britain 1700–1950 project (DEMLANG), funded by the Academy of Finland 2016–2020. The project PI is Päivi Pahta and the senior researchers are Turo Hiltunen, Minna Nevala, Arja Nurmi, Minna Palander-Collin, and Jukka Tyrkkö. [Go back up]
Allen, Wendy. 2007. “Australian political discourse: Pronominal choice in campaign speeches”. Selected Papers from the 2006 Conference of the Australian Linguistic Society, ed. by Mushin Ilana and Mary Laughren http://espace.library.uq.edu.au/view/UQ:12794
Androtsopoulos, Jannis. 2014. “Mediatization and sociolinguistic change. Key concepts, research traditions, open issues”. Mediatization and Sociolinguistic Change. Linguae & Litterae, vol. 36, ed. by Jannis Androtsopoulos, 3–49. Berlin and New York: De Gruyter.
Baker, Paul, Costas Gabrielatos, Majid Khosravinik, Michał Krzyżanowski, Tony McEnery & Ruth Wodak. 2008. “A useful methodological synergy? Combining critical discourse analysis and corpus linguistics to examine discourses of refugees and asylym seekers in the UK press”. Discourse and Society 19(3): 273–305.
Bello, Umar. 2013. “‘If I could make it, you too can make it!’ Personal pronouns in political discourse: A CDA of president Jonathan’s presidential declaration speech”. International Journal of English Linguistics 3(6): 84–96.
Chimbarange, Advice, Prosper Takavarasha & Francisca Kombe. 2013. “A critical discourse analysis of president Mugabe’s 2002 address to the World Summit on Sustainable Development”. International Journal of Humanities and Social Science3(9): 277–288.
Esser, Frank & Jesper Strömbäck. 2014. “Mediatization of politics: Towards a theoretical framework”. Mediatization of Politics. Understanding the Transformation of Western Democracies, ed. by Frank Esser & Jesper Strömbäck, 3–28. London and New York: Palgrave Macmillan.
Fetzer, Anita & Peter Bull. 2008. “‘Well, I answer it by simply inviting you to look at the evidence’. The strategic use of pronouns in political interviews”. Journal of Language and Politics 7(2): 271–289
Kovalyova, Natalia. 2005. “Presidential inaugural addresses: A study in a genre development”. Opening Windows on Texts and Discourses of the Past, ed. by Janne Skaffari, Matti Peikola, Ruth Carroll, Risto Hiltunen & Brita Wårvik, 39–53. Amsterdam: John Benjamins.
Moberg, Ulla & Göran Eriksson. 2013. “Managing ideological differences in joint political press conferences. A study of the strategic use of the personal pronoun ‘we’”. Journal of Language and Politics 12(3): 315–334.
Small Corpus of Political Speeches. 2007–. Compiled by Jukka Tyrkkö at the Universities of Helsinki and Tampere with the assistance of Veera Saarimäki. http://www.helsinki.fi/varieng/CoRD/corpora/SCPS/