Semantic text analysis and the measurement of ideological developments within fledgling democracies

Varieties of semantic grammars

Whereas thematic text analysis concentrates on word or phrase counts, a semantic text analysis examines theme relations. An example of thematic analysis is found in Laver and Garry’s (2000) investigation of differences between political parties. They analyzed party manifestos within which ‘desirable state policies’ are portrayed. In their study the investigators classify words and phrases used in the manifestos into numerous complex categories of state policies. For example, their category labeled Economy/+State+/ Budget/Spending/Health is applied to statements in which a desire is indicated for an increase of public spending in the field of health. Alternative categories were developed for reduced spending on health, increased spending on housing, younger eligibility for social security, etc. To cover the whole field of issues that play a role in the left–right classification, the investigators needed a large number of complex categories.

In a semantic text analysis, aspects of categories (e.g. state policies) are differentiated in accordance with the fields of a semantic grammar. For example, in their study of the cahiers de doléances of 1789, Shapiro and Markoff (1998) classified grievances in accordance with a Verb–Object semantic grammar within which members of a wide range of French organizations expressed their appeals that King Louis XVI might take a variety of actions (e.g. abolish, improve, modify, re-establish, etc.) regarding a variety of objects (e.g. the government, religion, the economy, etc.). Thus, instead of developing categories that comprise all the possible characteristics of one’s phenomena of interest (here, public grievances just prior to the French Revolution), one links categories in accordance with fixed semantic relations (those of verb to object, for instance). Koopmans and Erbe (2004) have done comparable research, only this time focusing on claims addressed to institutions or entire states within the European Union.

Most applications of semantic text analysis involve coding clauses according to the S–V–O triplet. Generally the unit encoded in accordance with such a semantic grammar is the grammatical clause (i.e. the sentence or part of a sentence that explicitly or impli-citly contains an inflected verb, an optional subject and/or object, plus all modifiers related to this verb, subject, and object). Yet such a semantic grammar might easily be expanded to include four syntactic components:

Because positions can only be taken by intentional agents, the agent cannot be any arbitrary subject. Instead it can only be a person or institution (comprising persons) and not a ‘thing’. The position commonly involves the use of a modal auxiliary verb. For example, an editorial that states ‘Our politicians ought to cooperate more’, would be encoded as politicians (agent) ought (position) to cooperate (action) with politicians (object). Other examples are Poles (agency) can (position) improve (action) their standard of living (object), and gasoline taxes (object) must (position) be lowered (action). (Note that in passive voice this last sentence specifies no agency, because its author did not explicitly implicate the government as the raiser of taxes.)

Table 1 illustrates the type of data matrix produced when texts are semantically encoded. Note that the numbers in the cells of this matrix correspond to specific concepts. For example, ‘Politicians ought to cooperate more’ could be encoded under the first ID using 12 as the code for ‘politician’, 7 for ‘ought’, and 34 for ‘cooperate’. Such semantically encoded data can be used to make inferences about the conditions under which texts’ authors take specific positions on others’ intentions. Accordingly, if a government’s totalitarian leadership were to initiate a public-relations campaign to ‘democratize’ its image in the US press, systematically encoded data could be used to test whether a significant increase had occurred from before to after the campaign in US news stories’ depictions of the regime (agency) attempting (position) to heed (action) its citizens (object).

OPEN IN VIEWER

Table 1.

A data matrix for a semantic text analysis.

ID-number	Agent	Position	Action	Object
1	12	07	34	12
2	08	03	22	88
3	14	03	34	70
4	10	01	35	70
5	16	11	29	52

The move from a thematic to a semantic text analysis expands the types of questions a researcher can answer. Referring to propaganda techniques in making this point, Roberts (1989: 169) notes that, in a thematic analysis, a possible research question would be: ‘What themes are mentioned in propaganda that are not mentioned in other communications?’ Using the semantic approach the question can be extended to: ‘What syntactic strategies are used by political leaders when their policies fail (or succeed)?’ Unlike the former question, the latter asks about concrete relations among concepts used within different social contexts.

Modality analysis

Generally speaking, modal auxiliary verbs are verbs that are used with (usually, the infinitive form of) another verb to express possibility, inevitability, impossibility, or contingency. In each modal usage there are two verbs associated with the subject, namely a modal auxiliary verb (e.g. want, hope, ought, refuse) and a main verb in infinitive form (typically an action). These usages are not intended to convey facts or to describe events; they are used to communicate something about the likelihood of the agent–action–object link. A semantic text analysis that investigates such uses of modal auxiliary verbs is called a ‘modality analysis’ (Roberts et al., 2010).

A modal clause is recognizable whenever it conveys intentionality in a way that can be transformed (in a manner agreeable to a native speaker) to a form that includes a modal auxiliary verb. For example, English speakers frequently use ‘have to’ instead of ‘must’ when indicating compulsion despite one’s intentions. On the other hand, modal auxiliary verbs are sometimes used in ways entirely unrelated to modality. Probably the best example here is the tendency among English speakers to convey future possibility rather than permission in their usage of the modal ‘may’. In modality analysis, the coder’s challenge is to capture how a text’s author understands others’ motivations (thereby getting into the mind of someone who is getting into someone else’s mind, as it were). ¹ Because modal auxiliary verbs convey intentionality, they can be used to learn about people’s motivations, their characterizations of each other, and thus about ideological shifts between and within their societies.

The semantic grammar used in a modality analysis always has two parts at its core. There is a modal form (indicating possibility, impossibility, inevitability, or contingency) and an associated rationale (e.g. expressed in liberal or social-democratic terms). Let us begin with a discussion of the four modal forms.

The modal form of ‘possible’ conveys to the reader that the modal’s predicate is an option for the modal’s subject. The modal form of ‘impossible’ conveys that the modal’s predicate is not an option for the modal’s subject. The form of ‘inevitable’ conveys that the subject–predicate link is imminent. Finally, the modal form of ‘contingent’ conveys the non-immanence of the subject–predicate link. These modals can be used to understand human motives during interactions, and to distinguish subtle nuances in discourse. For example, a sentence like ‘I can do it’ is coded as ‘possible’, implying capability (and, most likely, a sense of self-confidence); and a sentence like ‘I am not able to do it’ is coded as ‘impossible’, indicating incapability (and a potential sense of futility). A sentence like ‘she must be a good person’ is coded as ‘inevitable’, expressing certainty; and a sentence like ‘she is able not to do it’ is coded as ‘contingent’, indicating alternatives (and potential avoidance). Double negative expressions such as ‘not can not’ and ‘not must not’ are extremely rare in vernacular text.

Illustrations of the four modal forms are as follows:

Through modal usage a text’s source (i.e. its author or speaker) socially constructs what constitutes the possible, the impossible, the inevitable, and the contingent regarding the agent–action–object link. In addition, it is always reasonable for the source of a modal clause to be queried as to the rationale (or explanation of ‘why’) the agent is able, required, permitted, etc., regarding the clause’s predicate. For example, a politician might follow his statement, ‘We had to impose austere economic measures’, with the rationale, ‘otherwise our economy would have stagnated’. ²

Directions of democracy

Archival documents from the early 1990s afford a rich source of information on how Central and Eastern European nations transformed from totalitarian societies into democracies. In their texts on how democracy might develop in Central and Eastern Europe, both Ost (1993) and Comisso (1997) consider three socio-political orientations, namely (neo-)liberal, social-democratic, and populist/nationalist. Whereas Ost’s typology is predictive and primarily based on economic factors, Comisso’s emphasizes cultural-ideological factors. Nonetheless, the two typologies are in many ways mutually supplemental (see Touraine, 1991, for a more nuanced discussion of various definitions of democracy somewhat along the lines presented here).

By liberalism or neo-liberalism, both Ost and Comisso refer to the European sense of the term, indicating trust in the marketplace rather than in one’s government. The configuration of the market economy should take place as soon as possible. The once-planned economy is to be privatized. The establishment of a good foundation for investments is needed. As a special form of the liberal type, capitalism was the form most citizens desired immediately after 1989.

Comisso adds political ideas in her typology of the liberals. According to the liberals, the protection and augmentation of individual freedom is the fundament of a democracy. Liberals in Central-East European countries wished for transformations (especially capitalist reforms) to be implemented much faster than may have been realistic. In their view the changes proceeded too slowly.

Berlin’s (1969) concepts of negative and positive liberty help clarify the dual nature of liberalism’s emphasis on individual freedom. Negative liberty denotes the absence of obstacles, barriers, or constraints. Positive liberty refers to the possibility (or fact) of acting in such a way as to take control of one’s life and realize one’s fundamental purposes. Thus positive liberty requires negative liberty. In liberal terms, if one favors (positive) individual liberty one should place strong limitations on the (liberty-negative) activities of the state. Counter-arguments to Berlin’s view come from a broad range of non-liberal orientations. In brief, these critics argue that state intervention is often required for the successful pursuit of liberty, generally understood as individual or collective self-realization/self-determination.

For citizens with a populist/nationalist orientation changes occur too fast, and they may react by advocating an exchange of democracy for a strong, centralized government that will lessen the speed and thus the pain of reform. Such public reactions are often embedded in appeals to national pride, to ethnic rights, to xenophobic concerns, etc. Mann (2005) argues that widespread expressions of this sort are likely to foreshadow atrocities such as ethnic cleansing as the will of an ethnic majority takes precedence over the democratic process itself.

The social-democratic type shares with the liberals the idea that large-scale industry is of highest importance for the economy (Ost, 1993). However, social costs should remain minimal, making their position not only an economic one but a moral one as well. According to the social-democrats, the workers, who paid the costs for so long and who enabled a revolution, should have the opportunity to benefit now from the gains. They should not once again be sacrificed to the interests of future generations.

This approach strongly resembles Comisso’s second, egalitarian type. She also sees a resemblance between liberal and social-democratic types, both of which start from the individual. However, the difference with the liberal type is that the egalitarians do not endorse maximizing individual freedom. The egalitarian prefers an equal vote for everybody in the political process, even it entails state interference in the economy. This is contrary to the liberal democratic view, according to which the free-market sector should be unconstrained. For both Comisso and Ost, social-democratic and liberal orientations have reverse loci of trust, respectively with government and the market. Only with the former is a powerful government legitimate in ensuring equitable (re)distribution of the nation’s resources among its citizens.

Rather than conceiving of liberal, populist, social-democratic, and totalitarian as discrete orientations, Potůček (1999: 65) places liberal and totalitarian doctrines (communism, fascism, and so on) at opposite ends of a continuum. He differentiates points along this continuum in terms of support for values that would be expected of persons located at various points between the extremes. Potůček’s argument is that people’s value orientations along this continuum are important, because they determine possible ‘choices of society’. However, at the right end of his continuum, an awkwardly unexplained ‘leap’ exists from social-democrat to totalitarian – a leap that need not be made in studies of discourse within democratizing (i.e. no longer totalitarian) societies.

Within democratic (or democratizing) national contexts, a social-democratic versus liberal dialectic (with populism comprising an ever-present subversive avenue from democracy to totalitarianism) may afford more analytic purchase. Based on a review of the democratization literature, Roberts et al. (2009) note that it is typically via an alliance between the masses and the bourgeoisie that authoritarian regimes are ousted from power. Subsequent democratic efforts generally consist of struggles between parties that represent the respectively social-justice (or social-democratic) and market-justice (or liberal) interests of these two previously allied groups. Their subsequent modality analysis of post-1989 Hungarian editorials is based on the premise that legislative decisions generally call for choices between the policies advocated by representatives from each of these camps, because implementation of one of the former allies’ interests will commonly prove counter to the other ally’s interests:

And so, we now arrive at the familiar left–right political divide that characterizes party divisions within most contemporary democracies. On the left are those who justify government activities in terms of how well they manage a social system within which constituents’ hardships are minimized; on the right are those who justify such activities in terms of how well they ensure a stable economic field within which constituents can accumulate wealth. Beyond a consensus on the general benefits of a healthy national economy, at issue is whether the state’s role is to minimize its citizens’ risks or to maximize their opportunities. We now turn to a consideration of how modality analysis might be used to capture appeals to these two distinct ways of justifying political action.

Data

The data for this study are taken from a representative sample of editorials that appeared in the Hungarian newspaper, Népszabadság, between August 1990 (the month when Árpád Göncz became Hungary’s first post-communism elected president) and June 1997. ⁴ Népszabadság was selected because this newspaper is the least biased, yet consistent public forum for political and social discourse available on Hungary during the period of study. Our sample, consisting of 470 combinations of modal form and rationale, is described in detail in Roberts et al. (2009).

From this data set, 47 combinations were selected that illustrate liberal or social-democratic views in line with positions taken by Ost and Comisso. Two independently operating coders selected these combinations, taking into account the editorial context within which each was written. Details on the coding instructions and related issues are presented in Popping and Roberts (2009). Based on the 47 combinations that were assigned by at least one of the coders to either the liberal or social-democratic orientation, intercoder agreement following Scott (1955; Popping, 2010) is .88 (z = 31.16, p < .001). This subset is less than ideal, given that the texts were not sampled for the purpose of this study. Nonetheless, the subset is useful for illustrating how liberal and social-democratic rhetoric can be differentiated.

Measuring democracy-related ideological statements

Each modal statement-plus-rationale can in its most extended form be rendered according to the following template (or semantic grammar):

There is an R (rationale) why Z (action-object) is {possible / impossible / inevitable / contingent} for Y (agent) according to X (source).

Now the task is to identify R, X, Y, and Z. Here X is the author or speaker (presumably an adherent to a certain ideology), Y is the agent (to whom X believes the ideology applies), and Z is the predicate (the activity X is socially constructing as attributable to Y). Rationales (R) account for predicates’ possibility, impossibility, etc., in terms of political, economic, cultural, etc., ideological factors.

The speaker, X, and agent, Y, can be coded as liberal, socialist, and so on. Predicates, Z, can be coded in various degrees of detail. Yet it is the rationales that link the former three in accordance with ideological justifications. Roberts et al. (2009) distinguish among rationales’ political, economic, cultural, and security domains (see also Ben-Sira, 1977, 1979; Semetko & Valkenburg, 2000). They also argue that social-democratic orientations indicate necessity (e.g. what must responsibly be done). That is, predicates (i.e. expected or planned courses of events) are depicted as community members’ (inevitable) responsibilities. Accordingly, an illustration of a social-democratic orientation is as follows:

The government (lit: state) must support the weak [the poor people]. (10 September 1996) ⁵

Coded, it reads as: ‘The government (agent) must (position = inevitable) support (action) the weak (object).’ Supporting the poor is inevitable because of community members’ responsibility (here, via the government).

Liberal orientations typically convey possibility in the form of ability (e.g. of opportunities that can be pursued). Here a statement’s predicate is depicted as possible. If a predicate is impossible, it is because of the individual’s limitations. An illustration of a liberal view is the following:

Strong, young guys could work as porters (exclusively for tips only), they can, however, keep their numbered hats as souvenirs. (3 April 1991)

The first clause in the sentence is read as: ‘Strong, young guys (agent) could (position = possible) work as porters (action).’ The predicate in this illustration is depicted as possible. The rationale refers to the subject’s bodily resources.

Beyond characteristically social-democratic mentions of inevitability and liberal mentions of possibility, all other modal forms can be used in conveying social-democratic and liberal ideological orientations within public discourse. That is, we are not assuming that social-democratic modal rhetoric will always convey ‘inevitability’ and that liberal modal rhetoric will always convey ‘possibility’. Following Roberts (2008), all four modal forms are used in accordance with each political orientation. (See Roberts & Wang, 2010, for further illustrations along these lines in a comparative analysis of US and Japanese editorials or Roberts & Liu, 2014, for US and Canadian editorials.)

Unanticipated combinations of modal form and rationale

In analyzing modal use in other studies of such editorials, we have repeatedly found that references to possibility and inevitability are much more frequent than references to impossibility or contingency. The findings in Table 2 are no exception to this. The table contains the combinations of modal form and political orientation as found in the Hungarian data. For these data χ² = 23.86 (d.f. 6; p < .001), so the distribution of the frequencies over the cells given the marginals is far from random. Roberts et al. (2009) suggest that the modal form ‘possible’ is most prevalent among liberal modal statements and the ‘inevitable’ form is most frequent among social-democratic ones. As indicated in illustrations within the previous two subsections, one does find liberal statements indicating inevitability and social-democratic statements indicating possibility in Hungary’s post-1989 public discourse. However, they constitute a minority of the social-democratic and liberal modal statements in this discourse.

OPEN IN VIEWER

Table 2.

Modal forms identified as liberal and social-democratic.

	Liberal	Social-democrat	Neither	Total
Possible	14	5	234	253
Impossible	3	7	112	122
Inevitable	4	13	69	86
Contingent	0	1	8	9
Total	21	26	423	470

Let us consider a few cases inconsistent with our expectations, most notably modal statements in which economic reasons are given for inevitability and when welfare-related reasons are given for impossibility:

If we want to harvest to obtain a bigger share from the goods of the world economy, then we also have to do something. (10 December 1990)

This modal statement is from an editorial written by a trained economist, who follows a liberal line of thought, and therefore was jointly identified by our coders as expressing a liberal view. However, note that when its context is not taken into account, the statement might also have been made by a social-democrat (i.e. if the ‘harvesting’ were for our common wellbeing rather than for improved opportunities to individual entrepreneurs). Such ambiguity is greatest in situations where the ‘opposite’ of our theoretical expectations is found.

The modal form ‘impossible’ is used in combination with statements from either a liberal or a social-democratic perspective. These statements typically have the modal form ‘not can’. For the liberals, impossibility typically conveys the subject’s personal limitations (see illustration 2b above). The following typifies one of seven social-democratic statements of impossibility:

[The government] cannot leave local governments alone, the local authorities are unable to solve these problems due to the lack of money [lit.: materials]. (23 November 1992)

This statement might be considered as another way of expressing a parallel social-democratic modal statement that we would expect to find:

[The government] is obliged to support local governments, the local authorities are unable to solve these problems due to the lack of money [lit.: materials].

The reasoning is that ‘leaving local governments alone’ is logically equivalent to ‘not getting involved with (supporting?) local governments’, and that ‘not able (i.e. cannot)’ is logically equivalent to ‘must not’ – each conveying impossibility. Taking these together we get the former statement transformed into ‘The government must not not get involved with local governments’, which is equivalent to ‘The government must (or “is obliged to”) get involved with local governments.’ This reasoning is based on the rules of modal and first-order logics.

This transformation is applicable to all seven modal statements of impossibility from Table 2 that have a social-democratic orientation, justifying our moving them from the row for ‘impossible’ to the row for the modal form of ‘inevitable’. The resulting χ² for the data in the revised table now becomes 60.94 (d.f. 6, p < .001), thereby strengthening our argument that social-democratic public discourse is characterized by references to modal statements intended to convey inevitability.

The data used in the study on Hungarian public discourse were not collected with the goal of finding modal statements expressing liberal or social-democratic orientations. For that reason only 10% of all modal statements in the data set are useful for the present study. But if data had been collected with the goal of sampling modal rhetoric characteristic of these two orientations, we hypothesize that a pattern would have been found similar to that in (the revised) Table 2.

Conclusion

This article builds on a technique of semantic text analysis using modal forms. It is argued that in political discourse social-democrats tend to use the modal form ‘inevitable’, whereas liberals tend to use the form ‘possible’. It is shown that statements having a social-democratic rationale indeed generally contain the modal form ‘inevitable’ and that statements having a liberal rationale generally use the modal form ‘possible’.

So in studies of ideological developments in the direction of democracy, the modal form used within the nation’s editorials may be useful in predicting whether the direction of democratization is toward liberalism or social-democracy. Accordingly, the type of modality analysis performed here may be usefully applied to editorials and other archival texts from democratizing countries within which it is not completely clear what the country’s ideological direction might be. It affords an alternative measure of popular political perspectives that may not be in line with voting behaviors – a measure less susceptible to short-term interpersonal and historical influences.

However, this is only a beginning. To draw conclusions about the character of politic rhetoric, a larger, more representative sample needs to be used, necessarily including multiple newspapers from more than one new democracy. It should also be clear that the press systems in fledgling democracies are the voices of political positions.