1. Introduction
Why are some laws vaguely formulated, while others are concrete and detailed? Studies show that policymakers formulate legislative texts vaguely to reach compromise (Staton and Vanberg, Reference Staton and Vanberg2008; Hansen, Reference Hansen2016; Mousmouti, Reference Mousmouti2019). Vague law formulation also enhances implementation flexibility and adaptability to new circumstances (Li, Reference Li2017; Williams, Reference Williams2018; Li, Reference Li2019). However, vague law formulation may further result in unintended legislative output and reduce policy effectiveness and application uniformity (Williams, Reference Williams2018; Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024). Despite these important implications, we know surprisingly little about when policymakers advance the adoption of vague or concrete language in the formulation of legislative texts.
To address this gap, we study the formulation of laws in the European Union (EU) bicameral legislature. Specifically, we assess the use of vague versus concrete language in the initial positions of the European Parliament (EP) on legislative proposals.Footnote 1 While scholars have analyzed the substantive policy positions of the EP on specific issue dimensions (Costello and Thomson, Reference Costello and Thomson2010; Rasmussen and Reh, Reference Rasmussen and Reh2013; Costello, Reference Costello2022), their linguistic features remain underexplored.
Building on studies of EP success in bicameral negotiations with the second EU legislative chamber, i.e. Council of the European Union (henceforth, the Council) (Costello and Thomson, Reference Costello and Thomson2011, Reference Costello and Thomson2013; Kreppel, Reference Kreppel2018; Haag, Reference Haag2022) and its public responsiveness (Lo, Reference Lo2013; Sorace, Reference Sorace2022), we argue that the EP uses vague language strategically to shape EU policies and maintain its reputation as the institution representing directly the interests of EU citizens (TEU, Article 10). Based on two alternative assumptions about the EP’s main underlying policy goals, we derive contrasting expectations about its law-making strategies. On the one hand, if the EP aims to avoid legislative gridlock and to reach policy compromise, high levels of institutional conflict should prompt it to adopt more flexible and intentionally vague positions that a wider range of legislators can identify with and endorse. On the other hand, if the EP seeks to assert its position as strongly as possible in the policies ultimately adopted and implemented, institutional conflict should instead push it to take more concrete stances that are difficult to reinterpret or deviate from. Finally, regardless of whether the EP prioritizes compromise or specific policy outcomes, its core mandate is to represent EU citizens. To preserve a broad public appeal and avoid backlash, the EP is likely to tread carefully and blur its positions when public opinion over EU policy is divided or when the policy it advances is contentious.
To test our expectations, we analyze the vagueness of the EP positions vis-a-vis legislative proposals initiated between 2009 and 2019 using a dictionary-based automated content analysis. We used electoral manifesto data (Volkens et al., Reference Volkens, Burst, Krause, Lehmann, Matthieß, Merz, Regel, Weßels and Zehnter2020) to measure the intra- and inter-institutional conflicts and Eurobarometer surveys to derive indicators of public opinion division over EU policy integration (Ershova et al., Reference Ershova, Yordanova and Khokhlova2024). To estimate the implications of EP amendments for EU authority, we relied on semi-supervised machine learning, following Yordanova et al. (Reference Yordanova, Ershova, Khokhlova, Obaid Ul-islam and Glavaš2025a).
The results show that the EP’s policy goals shape the language it uses to formulate its legislative positions. First, the EP adopts less vague stances when it has policy disagreements with the Council. This suggests that it prioritizes locking its own legislative position in adopted legislation and policy outcomes over reaching a legislative compromise. In other words, we do not find support for the rationale that strategic vagueness helps overcome institutional conflict to reach agreement. Instead, the EP becomes more concrete given institutional conflict to assert its positions and assure that policy-as-implemented remains close to the policy-as-adopted. Second, the Parliament advances vague legislative positions when EU citizens’ policy views get more divided and when it advances contentious policy proposals, and specifically proposals that would change the level of supranational authority, which has become increasingly politicized (Hagemann et al., Reference Hagemann, Hobolt and Wratil2017; Wratil, Reference Wratil2018; Börzel et al., Reference Börzel2023). The EP thus appears to use vague language strategically to avoid public backlash and, thus, preserve its legitimacy. Altogether, these findings demonstrate that the EP’s use of vague language in its position formulation depends on the type of (potential) conflict that confronts it as well as its primary objectives in terms of policy and representation.
These results enhance our understanding of the mechanisms driving strategic framing of legislative texts (Jegen and Mérand, Reference Jegen and Mérand2014; Williams, Reference Williams2018; Hurka et al., Reference Hurka, Rueß, Cowburn and Kaplaner2025). Furthermore, by accounting for the lexical choices in EP amendments prepared for legislative negotiations, we add to knowledge of the EP internal policymaking (Costello and Thomson, Reference Costello and Thomson2010; Rasmussen and Reh, Reference Rasmussen and Reh2013; Costello, Reference Costello2022) and inter-institutional bargaining in the EU (Cross and Hermansson, Reference Cross and Hermansson2017; Kreppel, Reference Kreppel2018; Broniecki, Reference Broniecki2020; Laloux and Delreux, Reference Laloux and Delreux2021; Haag, Reference Haag2022). Last, our results add to the literature on the role of public opinion in shaping the legislative behavior of the EP. Specifically, they show that its responsiveness to public opinion is not expressed only through substantive policy positions but also through how clearly or vaguely it formulates those positions.
2. Strategic vagueness in legislative text formulation
Legislative texts—proposals for new laws, amendments, or adopted acts—introduce, alter, revise, or clarify policy measures. Yet, as Williams (Reference Williams2018, p. 5) puts it, “Legislation increasingly does not, therefore, offer a determinate ruleof law. (Williams Reference Williams2018, p.5)”. This suggests that legislative texts are often formulated vaguely and lack precision (see also Endicott, Reference Endicott, Bhatia, Engberg, Gotti and Heller2005; Shuangling, Reference Li2017; Li, Reference Li2019).
Research on the indeterminacy in legislative texts identifies vague language as a source of ambiguity (Li, Reference Li2017, p. 100; Byers, Reference Byers2021) and imprecision (Hurka et al., Reference Hurka, Rueß, Cowburn and Kaplaner2025). Other sources of ambiguity identified in the literature on electoral campaigns include political parties de-emphasizing positions or being inconsistent on them (Lefevere, Reference Lefevere2024, p. 334). Deliberate avoidance or policy inaction is difficult to detect, while ambivalence through the combination of inconsistent policy measures is more likely to bring tangible and reputational costs rather than benefits for legislators. Thus, we focus on vagueness as a source of ambiguity that legislators are likely to employ strategically to achieve their goals.
Vague formulations allow for multiple interpretations and affect how laws are understood, implemented, and enforced. Thus, they may create uncertainty about obligations and permissible actions, which in turn obstructs proper law implementation (Xanthaki, Reference Xanthaki2014; Shuangling, Reference Li2017; Li, Reference Li2019; Byers, Reference Byers2021; Fowler Reference Fowler2023; Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024).
Despite these potential challenges, many instances of imprecision of legislative texts are not the result of poor drafting, but of a strategic choice (Williams, Reference Williams2018; Mousmouti, Reference Mousmouti2019; Hurka et al., Reference Hurka, Rueß, Cowburn and Kaplaner2025). Specifically, policymakers regularly resort to the strategy of constructive ambiguity, i.e. intentionally introducing vague and ambiguous language, thus blurring the scope and methods of law application (Jegen and Mérand, Reference Jegen and Mérand2014; Byers, Reference Byers2021). The ambiguity is constructive when it allows moving the policymaking forward and gives interpretive leeway that can be used to advance political goals. Studies analyzing the language characteristics in legal texts identify several motives behind indeterminate formulations.
First, some contributions highlight that lack of clarity may facilitate law adoption and contribute to policy change. Mousmouti (Reference Mousmouti2019, p. 49) suggests that vague formulation of policy measures is a practical solution that allows to bridge political divides. It helps to avoid conflicts among policymakers or their relevant audiences while allowing them to nonetheless be seen as active decision-makers as opposed to not taking any policy action at all when anticipating conflict (Williams, Reference Williams2018, p. 19; see also Xanthaki, Reference Xanthaki2014, p. 87). For instance, in the case of the United Nations’ Security Council, Byers (Reference Byers2021) claims that the possibility of multiple interpretations provided by the imprecision of resolutions is a deliberate choice that facilitates compromise between negotiating parties. Countries can then choose an interpretation that aligns with their preferences (see also Fowler Reference Fowler2023, p. 1395). In a similar vein, Jegen and Mérand (Reference Jegen and Mérand2014) argue that in the EU context, the use of constructive ambiguity may prevent strong opposition from member states and accommodate their heterogeneous policy positions. Likewise, Hansen (Reference Hansen2016) underlines that ambiguity can help prevent negotiation failures arising from divergent actors’ preferences on new measures in the EU arms export control regime.Footnote 2
Second, the adoption of less determinate laws may reduce potential non-compliance (Hansen, Reference Hansen2016). Studies show that less concrete language gives implementing actors greater discretion (Xanthaki, Reference Xanthaki2014, p. 88; Mousmouti, Reference Mousmouti2019), thereby enabling more tailored policy implementation (Williams, Reference Williams2018). Furthermore, it provides flexibility in case of unforeseen circumstances (Li, Reference Li2017; Williams, Reference Williams2018, p. 12; Li, Reference Li2019). Less precise texts allow for more leeway and prompt adjustments if it becomes necessary to address a pressing issue or adapt to a changing political situation. However, other studies suggest the opposite. Vaguely formulated laws may decrease the legislation quality and be less effective due to problems with their interpretation and, consequently, correct and timely implementation (Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024; Hurka et al., Reference Hurka, Rueß, Cowburn and Kaplaner2025). In the EU context specifically, Osnabrügge and Vannoni (Reference Osnabrügge and Vannoni2024) show that vagueness tends to reduce the probability of member states compliance with supranational policy measures.
Finally, based on the literature on electoral campaigns, political parties adopt vaguer positions when they seek to accommodate a wider range of public views (Nasr, Reference Nasr2023; Lefevere, Reference Lefevere2024). Once in parliament, when public opinion is divided, legislators reduce the concreteness of their speeches if they concern matters that are salient to the citizens (Khokhlova, Reference Khokhlova2025). Legislators could conceivably employ a similar strategy in lawmaking and resort to the use of vague language as a way to increase their appeal to wider audiences. Specifically, by adopting laws with imprecise formulations that allow for flexibility in implementation, policymakers may preempt public backlash in cases where the introduced policy measures do not align with citizens’ views or where citizens and other relevant stakeholders hold competing preferences.
In summary, this literature shows that policymakers have incentives to reduce law concreteness to accommodate diverging preferences and secure adoption, flexible implementation, and adaptability to changing circumstances, but at the risk of reduced policy effectiveness and improper implementation. Given these trade-offs, what factors motivate legislators to adopt this strategy despite its potential negative effects?
To address this question, we focus on the case of the EP. We study its proposed amendments to draft EU laws at the start of the inter-institutional legislative negotiations with the Commission and the EU Council. This allows to analyze the linguistic properties of initial legislative positions that precede any concessions made during negotiation process. The next section details how the EP adopts its policy positions, after which we present our theoretical arguments.
3. The EP’s policy positions for legislative negotiations
Nowadays, the vast majority of EU laws are adopted under the ordinary legislative procedure (OLP), formerly known as co-decision. Under this procedure, the European Commission drafts legislative proposals, which the Council and the EP can amend and adopt at any of up to three readings or reject. Yet, most of the dossiers are adopted at first reading following an agreement reached in informal negotiations between the three institutions in secluded meetings, so-called trialogues.
Prior to the first trialogue, the EP and the Council formulate their policy positions in the form of amendments to the Commission’s proposals, which serve as mandates of their representatives for trialogue negotiations. Following Rule 72 in the EP Rules of Procedure (Parliament, Reference Parliament2023), the negotiation position of the Parliament is based on a legislative report drafted by a rapporteur and adopted by the responsible parliamentary committee. If the EP plenary does not object to it, this report serves as the EP negotiation mandate for its negotiation team in trialogue meetings.Footnote 3 On the contrary, the amendments to the committee report can be made in plenary by a set deadline, only after which it can serve as EP mandate and allow for entry into trialogue negotiations. This procedure for entry into inter-institutional negotiations is relatively new. Up until a change of the EP Rules of Procedures in late 2012 (Parliament, Reference Parliament2012), a parliamentary delegation, led by the rapporteur, could enter into trialogue talks without an explicit EP mandate. This resulted in the final committee report fully taking over and reflecting rather than informing the trialogue deal (Yordanova, Reference Yordanova, Muller-rommel and Bertoa2015).
The extent to which committee reports and the EP representations in trialogues reflect the overall positions of the Parliament has been a subject to scientific debate.Footnote 4 However, we still know little about the linguistic choices that underpin the parliamentary starting negotiation positions. The EP can adopt vaguely formulated amendments that help it to limit opposition to its positions by leaving room for interpretation, but which also reduce clarity on the applicability and lead to deviations of policy-as-implemented from policy-as-adopted. Alternatively, it can put forward concrete amendments, which increase the precision of suggested measures and reduce potential interpretation problems but also, by eliminating interpretative leeway, risk provoking opposition. Assuming this choice is strategic, the central question of this paper is: Under what conditions does the EP employ vague language in its amendments to Commission proposals? To assess it, we draw on the literature on legislative negotiations in the EU and EP responsiveness.
4. Strategic vagueness by the European Parliament
We argue that when formulating its positions on proposed policy measures, the EP may rely on a strategy of constructive ambiguity (Jegen and Mérand, Reference Jegen and Mérand2014; Byers, Reference Byers2021). This term was originally applied to the EU to explain member states’ negotiations and decision-making in the Council. We hold that the EP also uses this strategy to advance its goals: shaping policy and serving the EU citizens. Specifically, we argue that, on the one hand, the EP’s strategic use of vagueness in policy formulation depends on whether it is primarily driven by the goal of reaching a policy compromise or asserting its own position in adopted legislation and policy outcomes. On the other hand, it depends on the public opinion constraints it may be subject to. We develop these arguments below.
4.1. Reaching legislative compromise
When the EP’s priority is to avoid legislative deadlock and secure policy compromise, institutional conflict—be it conflict within the EP or between the EP and the Council—incentivizes it to increase the vagueness of its amendments, and thus the flexibility and room for their interpretation, as a way to facilitate finding agreement.
First, we expect intra-institutional conflict, i.e. the divergence of policy preferences within the EP, to be crucial for its legislative position formulation. It is difficult for a highly internally divided Parliament to agree on a common position. To find internal agreement when its members hold more divergent preferences, we expect the EP to use less concrete language in the formulation of its legislative positions. Studies show that less precise policy measures help accommodate conflicting preferences (Jegen and Mérand, Reference Jegen and Mérand2014; Hansen, Reference Hansen2016; Williams, Reference Williams2018; Mousmouti, Reference Mousmouti2019; Byers, Reference Byers2021). Thus, by formulating positions vaguely, rapporteurs may be able to pacify parties with conflicting preferences (Mousmouti, Reference Mousmouti2019, p. 49) and facilitate mandate adoption for upcoming trialogue negotiations. Drawing on this, we hypothesize:
H1a (EP internal division): The higher the level of intra-institutional conflict in a policy domain, the more vague are the EP amendments to the proposed legislation in that domain.
Second, inter-institutional conflict provides incentives for more vague, and thus flexible, position formulation to help find a common ground with the Council. Based on the literature on inter-institutional negotiations, the EP’s legislative influence is greater when its policy positions are congruent with the Council’s position, and vice versa (Kreppel, Reference Kreppel2018). Haag (Reference Haag2022, p. 344) shows that the EP’s relative ability to exert influence on the legislative output of trialogue negotiations is stronger when its own and the Council’s positions in trialogue negotiations converge. This suggests that when the EP’s and the Council’s policy positions are more aligned, the EP can introduce more concrete amendments to the Commission’s proposals in anticipation that the Council will approve them. In contrast, when the Council and EP hold divergent positions, the EP needs to compromise if it wants to avoid legislative gridlock. It then has incentives to adopt a less determinate stance as vague amendments leave interpretive leeway (Byers, Reference Byers2021; Fowler Reference Fowler2023) that can preempt active opposition from the Council, thus facilitating an inter-institutional agreement that reflects the EP’s position. Therefore, we expect that:
H1b (EP-Council conflict): The higher the level of inter-institutional conflict in a policy domain, the more vague are the EP amendments to the proposed legislation in that domain.
4.2. Shaping legislative output and policy outcomes
Instead, when the EP mainly seeks to assert its specific positions in the final legislative output and policy outcome rather than preventing legislative gridlock per se, heightened institutional conflict rather creates incentives for drafting more concrete, and thus precise and unambiguous amendments. Clear formulations restrict the scope for reinterpretation during negotiations and implementation, which intra-EP and EP-Council divisions predispose to, thus helping the Parliament safeguard its preferred outcomes and ensure their proper implementation.
First, when policy conflict in the EP is high, intra-EP agreements are fragile and can break if inter-institutional deals diverge from the EP’s position. This means the EP delegation needs to avoid deviation in trialogues, as divergent deals will not pass on the parliamentary floor. This constraint strengthens the EP’s bargaining position, since significant changes to its stance would not be approved, allowing it to extract concessions from the Council (Haag, Reference Haag2022). This enables the EP to use more precise language in its amendments without fearing that the Council will block them. We can therefore expect the EP delegation to uphold a clear, narrowly defined position when it reflects a narrow parliamentary compromise:
H2a (EP internal division): The higher the level of intra-institutional conflict in a policy domain, the less vague are the EP amendments to the proposed legislation in that domain.
Second, when a conflict with the Council is high, there are good reasons for the EP to expect the policy-as-implemented to deviate from the policy-as-adopted. By leaving a lot of room for interpretation, vague language in legislation can reduce policy effectiveness and the uniformity of law implementation and application (Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024). In the EU institutional context, this is particularly likely because member state governments, represented by the Council, are not just co-legislators but also the ones who get to implement EU policies. When these policies are vaguely formulated, governments have more flexibility to interpret them differently than intended by the legislators, which is especially likely if the EP and Council disagreed when coining the policy to begin with. Hence, in the face of diverging views with the Council, to safeguard that policy application is in line with its policy positions, the EP should seek to increase precision of policy measures and limit interpretive leeway. This leads us to the opposite expectation, namely:
H2b (EP-Council conflict): The higher the level of inter-institutional conflict in a policy domain, the less vague are the EP amendments to the proposed legislation in that domain.
4.3. Avoiding public backlash
Furthermore, we argue that public opinion constraints may drive the formulation of more vague EP policy positions. This argument is rooted in the observation by Byers (Reference Byers2021) that policymakers respond to public constraints when making ambiguous agreements on the international level as well as in the evidence that the use of vague language helps appeal to and bridge the different preferences of diverse publics in electoral contexts (Nasr, Reference Nasr2023). Specifically, the EP is mandated by the Treaty of the EU (Article 10) to represent the citizens’ views in the law-making process and, thus, needs to respond to public opinion when legislating. Research shows that the EP’s involvement in the legislative process contributes to the better reflection of citizens’ preferences in policy outputs (Sorace, Reference Sorace2022), supporting its reputation as representative of the public. Consequently, when formulating policy positions that shape EU laws, the EP must consider both the dynamics of public opinion and how the public will perceive and evaluate adopted laws, i.e. their potential effects on future public opinion.
In the EU, the very authority of supranational institutions and further integration have become increasingly politicized and divisive, with divergent views on the extent to which the EU rather than national authorities should have a say in specific policies (Hooghe and Marks, Reference Hooghe and Marks2009). This has reduced the opaqueness of choices on the supranational arena and incentivized policymakers, who no longer enjoy citizens’ permissive consensus for deeper integration, to factor in public opinion when defining EU authority (De Bruycker, Reference De Bruycker2019; Wratil, Reference Wratil2019). Consequently, the importance of the integration dimension of political contestation and public views along this dimension has grown considerably in EU policymaking (Hagemann et al., Reference Hagemann, Hobolt and Wratil2017; Wratil, Reference Wratil2018; Börzel et al., Reference Börzel2023) and in the EP, specifically (Otjes and van der Veer, Reference Otjes and van der Veer2016; Blumenau and Lauderdale, Reference Blumenau and Lauderdale2018).
EU institutions have responded to EU politicization in different ways. In the Council, ministers tend to vote against proposals that expand EU authority when they face Eurosceptic publics back home (Hagemann et al., Reference Hagemann, Hobolt and Wratil2017). Furthermore, Europhile governments tend to accommodate each other’s domestic constraints stemming from Eurosceptic public opinion (Mariano and Schneider, Reference Mariano and Schneider2022). Instead, the Commission typically holds and has maintained a pro-integration stance because its competences depend on EU authority, incentivizing it to preserve or seek deeper integration. For instance, the Commission virtually never proposes policies decreasing EU authority (Yordanova et al., Reference Yordanova, Khokhlova, Ershova, Schmidt and Glavaš2025b). Conversely, it tends to capitalize on public pressures for EU action by prioritizing pro-integration files when a policy is salient for the citizens (Koop et al., Reference Koop, Reh and Bressanelli2020). The EP and the Council give less discretion to the Commission for policy implementation when public support for EU policy action is lower (Ershova et al., Reference Ershova, Yordanova and Khokhlova2024). The integration dimension also increasingly structures voting in the Parliament (Otjes and van der Veer, Reference Otjes and van der Veer2016; Blumenau and Lauderdale, Reference Blumenau and Lauderdale2018).
Building on this literature, we expect the EP to be more cautious about how it formulates its positions on proposed legislation in policy areas in which public divisions over EU policy integration are larger. In such areas, it may become challenging to craft a policy position that resonates with the public when public preferences grow more divided, and there is no clear majority opinion to represent. Formulating vague policy positions then can help the Parliament avoid the appearance of favoring certain preferences over others and, consequently, alienating a substantial part of its public. Furthermore, by allowing tailored and flexible implementation in the face of diverging preferences over supranational policy measures among EU citizens, the EP mitigates the risk of increasing public dissatisfaction with EU policies and undermining own legitimacy. This leads us to the following expectation:
H3 (Public opinion division): The more divided public opinion over EU policy action grows in a policy domain, the more vague are the EP amendments to the proposed legislation in that domain.
While EU involvement may be more contested in some policy areas than in others, legislation that explicitly modifies EU authority, i.e. the scope or level of supranational competences, is particularly contentious (Hagemann et al., Reference Hagemann, Hobolt and Wratil2017; Hooghe and Marks et al., Reference Hooghe, Marks, Lenz, Bezuijen, Ceka and Derderyan2017). Because of this, the EP is susceptible to public backlash when its amendments seek to alter EU authority. Employing vague language that allows for reinterpretation and adaptation to domestic contexts can help the EP limit potential audience costs associated with legislation that has implications for EU policy integration. Therefore, we further hypothesize that:
H4 (Contentious changes): The more the EP seeks to alter EU authority with its amendments to proposed legislation, the more vague are its amendments to that legislation.
5. Data and methods
To test our hypotheses, we compiled a dataset of 359 OLP procedures initiated and adopted between December 2009 and May 2019, for which we could retrieve the trialogue tables prepared by the EP and Council staff for the first inter-institutional legislative negotiation.Footnote 5 The trialogue tables split the text of proposals into provisions (quasi-paragraphs listed in rows) and, for each provision, present in four columns the corresponding positions of the Commission (proposal), the EP and the Council as well as the potential compromise text. Thus, these tables allow for an easy comparison of texts across the institutions.
We categorized the legislative procedures in our sample into the policy domains defined by the European Union Policy Agendas Project (CAP-EUPAP) (Alexandrova et al., Reference Alexandrova, Carammia, Princen and Timmermans2011) using the EuroVoc descriptors associated with each act, following Yordanova et al. (Reference Yordanova, Khokhlova, Ershova, Schmidt and Glavaš2025b) (see Appendix 1 for details). We also retrieved additional information (such as the date of the Commission’s proposal and legislative instrument) for each act from the EUR-LEX dataset.
5.1. Dependent variable: increase in vagueness
As our dependent variable (DV), we measure the vagueness of the EP position in the first trialogue negotiations relative to the vagueness of the original Commission’s proposal. We opted for such a relative rather than an absolute measure to account for the EU legislative process. As the EP is not the agenda-setter, its position is a response to and, thus, inherently affected by the Commission’s proposal and its baseline level of vagueness, which could be driven by unobserved factors unrelated to the EP’s strategic behavior. A relative measure allows us to isolate the EP’s contribution to changes in legislative precision, encapsulated in its amendments, rather than capturing variation driven by the Commission. Therefore, our DV offers a valid and reliable measurement of the extent to which the EP’s positions explicitly blur legislative proposals.
To construct the DV, we estimated the vagueness of legislative positions following the approach of Eichorst and Lin (Reference Eichorst and Lin2019), the measures of Commission’s and EP’s vagueness equal the percentage of vague minus the percentage of concrete terms.Footnote 6 However, while Eichorst and Lin (Reference Eichorst and Lin2019) rely on tentative and certain words of the English-language Linguistic Inquiry and Word Counts (LIWC) dictionaries, these proved to be less suitable for identifying vagueness in EU legislative texts. The majority of LIWC terms are context-dependent and may indicate either vagueness or precision depending on their usage in a sentence or provision.
Therefore, we adopt a more tailored and robust approach and define vagueness as imposing weaker obligations in legislative acts and precision as imposing stronger obligations.Footnote 7 That loosely relates to what Hurka et al. (Reference Hurka, Rueß, Cowburn and Kaplaner2025, p. 741) refer to as dilution, which indicates whether a provision is softened, weakened, or made discretionary with the use of vague words and expressions. We capture this by counting the occurrences of the following contrasting groups of terms: may, might, should, encouraged to, invited to, recommended to (vague) and must, shall, required to, obliged to, prohibited from (concrete). These terms correspond to commonly used legal expressions indicating lower and higher obligation strength in legislative drafting: the former signal greater discretion and offer interpretative leeway, whereas the latter impose strict mandatory requirements. Crucially for our DV, changes in word choice—for instance, from must to may or from obliged to to recommended to—are substantively meaningful. When the EP replaces must with may, it provides greater interpretive authority and flexibility to those responsible for law implementation, and vice versa.
First, for each provision within the Commission’s proposals and EP positions, we counted the number of vague and concrete terms.Footnote 8 Second, for each provision, we estimated the percentage of vague and concrete words and subtracted the latter from the former. The resulting differences indicate provision vagueness and vary between −18.182 and 20.000 for provisions in the Parliament’s position and between −25.000 and 20.000 for provisions in the Commission’s proposals. Lower values indicate more concrete language, meaning that terms strengthening obligations are used more frequently than those weakening them.
Third, to estimate to act-level vagueness of the Commission and EP positions, we took the average vagueness across all provisions in their respective versions of the actFootnote 9 (see Appendix 4, where we present densities of institutional position vagueness by policy area and EP term). As a last step to obtain our DV—Increase in vagueness—for all procedures in our dataset, we calculated the difference in the vagueness between the EP and the Commission texts. Larger values of this variable indicate more vague EP position relative to the Commission’s proposal, and vice versa. We summarize DV distribution in Figure 1 and discuss its validity in Appendix 5.
Distribution of the dependent variable.
5.2. Independent variable and controls
Our first independent variable (IV)—Intra-EP conflict—captures the degree of intra-institutional conflict in a specific policy area at the time of the first trialogue. To construct this variable, we relied on national parties’ electoral manifestos from Volkens et al. (Reference Volkens, Burst, Krause, Lehmann, Matthieß, Merz, Regel, Weßels and Zehnter2020). The reliance on the positions of national political parties rather than European party groups to measure intra-EP conflict is justified with the lack of electoral control by the latter in the EP. It also follows the convention in related research (Haag, Reference Haag2022). Furthermore, we use manifestos rather than available expert surveys of party positions because the latter is much more limited in terms of its temporal and policy area coverage. We matched the Manifesto policy categories with the 22 policy domains used for the classification of the legislative acts in our sample (see Appendix 5) and estimated the policy preferences of each national political party present in the 7th and the 8th parliamentary terms following the log-transformation approach of Lowe et al. (Reference Lowe, Benoit, Mikhaylov and Laver2011). Using these estimates, we then calculated the level of intra-EP conflict in each policy domain based on the polarization index proposed by Dalton (Reference Dalton2008).Footnote 10 Larger values of the index indicate a higher degree of the Parliament’s internal disagreement, and vice versa.
The second IV—EP-Council (EP-CN) conflict—is measured as the absolute distance between policy preferences of the co-legislators at the start of trialogues. For the construction of the EP policy preferences, we summed the policy preferences of all political parties in a given EP term obtained in the previous step, weighted by the parties’ parliamentary seat shares.Footnote 11 To calculate the Council’s position, we identified the composition of the national governments as recorded in the ParlGov database (Döring and Manow, Reference Döring and Manow2020). Next, following the same approach as described for parties in the EP, we obtained the measures of government parties’ preferences for each specific policy domain. We then estimated the mean positions of all member state governments between 2008 and 2019, weighing the positions of their constituent parties by their share of government seats in the national parliaments. The aggregate Council’s position was constructed by taking the mean of the governments’ positions, weighted by their respective countries’ population sizes. Finally, we measured inter-institutional conflict as the absolute difference between the preferences of the two institutions.
Our third IV measures public opinion division across EU member states and comes from Ershova et al. (Reference Ershova, Yordanova and Khokhlova2024). This variable is based on Eurobarometer questions, classified into CAP-EUPAP policy domains.Footnote 12 It measures the absolute difference between the shares of citizens supporting and opposing EU action in a given policy domain and year, which is subtracted from 1 so that “1” indicates a 50:50 split and “0” complete agreement of citizen on the need for more or less EU action in the policy domain of a proposal. To account for the baseline level of public opinion division to which the Commission may respond when formulating its proposals (Yordanova et al., Reference Yordanova, Khokhlova, Ershova, Schmidt and Glavaš2025b), and to isolate the constraints faced by the EP in its position formulation, we measure this variable as the change in public opinion division between the Commission proposal and the first trialogue meeting. Higher values of the resulting variable, Increase in Public opinion division, indicate that EU citizens’ opinions became more divided, while lower values indicate that they were more united at the time of the trialogue.
We operationalize the contentiousness of the position advanced by the EP with its implications for EU authority. For the purpose, we used information from the trialogue tables and assessed whether EP revisions implied more or less policy integration than envisioned in the Commission proposals. Specifically, we combined hand-coding and machine learning classification of the provisions containing Commission proposal text and corresponding EP text, using the approach and codebook developed and validated by Yordanova et al. (Reference Yordanova, Ershova, Khokhlova, Obaid Ul-islam and Glavaš2025a) to measure the Council’s position based on trialogue tables (for details, see Appendix 6). We classified provisions in trialogue tables containing the Commission proposal and corresponding EP texts. For manual annotation, we sub-sampled approximately 5 percent of provisions, balanced across policy domains and years. Using an extensive codebook, two coders independently labeled each provision as pro-status quo 0, pro-EU authority expansion 1, or containing only technical information with little relevance for EU authority 2. Inter-coder agreement for EP position annotations approximated 1, while for Commission position annotations it was 0.997. Based on these annotations, we trained a RoBERTa Large classifier (Liu et al., Reference Liu, Ott, Goyal, Du, Joshi, Chen, Levy, Lewis, Zettlemoyer and Stoyanov2019) and applied it to remaining provisions in our dataset, generating probabilities for each provision falling into these three categories.
To measure the institutional positions on the level of legislative acts, we calculated the mean estimated probabilities of expanding the EU authority across all provisions within the Commission’s and EP’s trialogue columns, respectively. We measured our IV, Increase in EU authority, as the position of the EP minus the position of the Commission, so that higher values indicate that the EP’s position on a legislative proposal is more integrationist (more likely to expand EU authority) than that of the Commission. Furthermore, we included an additional quadratic term of this IV in our analysis to assess the hypothesized non-linear effect of this variable.
We controlled for several additional factors that have a potential to shape the Parliament’s use of vague and concrete language in its policy positions. First, following the literature on EU politicization (De Bruycker, Reference De Bruycker2019), we included two variables that capture further forms of public constraints to which the EP might be sensitive. This includes Public salience and Societal actor mobilization in the policy domain at the time of the trialogue. To measure public salience, we followed Wratil (Reference Wratil2019) and Gilens (Reference Gilens2012) and used the share of Don’t know (DK) responses relative to the total number of meaningful responses to the sampled Eurobarometer questions, subtracting the resulting value from one. This variable captures the relative citizen awareness of (potential) EU policy, serving as a proxy for their interest in it.Footnote 13 Our measure of Societal actor mobilization is based on the European Commission’s Transparency Register, which contains information on the number, type, and policy focus of organizations from all EU member states, which represent diverse opinions and can be consulted during supranational policymaking. It sums the number of organizations registered across EU member states in each year and policy area. Business organizations are excluded as they are more likely to advance their own interests rather than transmit citizens’ preferences (for more details, see Ershova et al., Reference Ershova, Yordanova and Khokhlova2024). Arguably, the mobilization of variety of actors representing alternative citizens’ views makes it more challenging for the EP to accommodate these alternative views during the legislative process, which in turn can incentivize it to use less concrete language as a way of limiting audience costs. Missing values for these variables are imputed using a mean imputation approach.
Second, we controlled for intra-EP policy salience. To construct this variable, we summed up the shares of negative and positive policy references in national parties’ electoral manifestos (see Appendix 6), capturing their emphasis on CAP-EUPAP policy domains. Specifically, for each legislative act in our sample, Intra-EP salience was calculated as the average salience that national parties represented in the EP at the time of the first trialogue attached to the policy area of the act.
Third, we accounted for the characteristics of legislative acts. This included the legislative instrument—Directives, Regulations, or Decisions (the reference category)—based on EUR-LEX. We expected Directives, which must be transposed or incorporated into the national legislation before they enter into force, to be formulated more vaguely to give member states discretion during implementation. Furthermore, we controlled for the Complexity of legislation, as more complex laws that touch on multiple policy domains might be strategically formulated in vaguer terms to find an agreement. We constructed the measure by counting the number of EuroVoc descriptors associated with each procedure, as suggested by Migliorati (Reference Migliorati2021).
Fourth, we controlled for temporal effects. This included the number of Months from proposal to trialogue. The longer this time is, the more time the Parliament has to agree on more concrete or substantial amendments internally. Last, we controlled for the Months to the EP elections, i.e. the number of months between the dates of the first trialogue and the subsequent EP elections. Several studies suggest that intra-EP politics is shaped by the supranational electoral cycles (Lindstädt et al., Reference Lindstädt, Slapin and Wielen2011; Koop et al., Reference Koop, Reh and Bressanelli2018). Electoral proximity intensifies scrutiny of EP policymaking, strengthening incentives to avoid institutional push-back and audience costs, which vaguer parliamentary amendments can facilitate.
6. Results
Given the continuous DV, we tested our hypotheses relying on ordinary linear regression. We estimated two model specifications (Table 1), Model 1 including only IVs, and Model 2 including all controls. We note that this analysis employs observational data, which allows to establish correlational but not causal relationships.
OLS regression results
Table 1 Long description
Two OLS models estimate the relationship between political and institutional factors and the increase in vagueness between the Commission proposal and the EP policy position, using 352 observations in each model.. Across both models, the coefficient for EP-CN conflict is negative and statistically significant, with a larger magnitude in model 2. Increasing public opinion division is positively associated with increase in vagueness between the proposal and the EP position in both models. The linear EU authority term is not statistically significant, but the squared EU authority term is positive and statistically significant in both models, suggesting a non-linear relationship. In model 2, public salience is weakly negative, while societal actor mobilization, intra-EP salience, complexity, timing variables, and legal form indicators show small, statistically uncertain associations. Model fit improves modestly when controls are added, with R-squared rising from about 0.05 to about 0.09 and residual variation declining slightly. As these estimates are based on observational regressions, they should be interpreted as associations rather than causal effects.
Note: + p < 0.1; * p < 0.05; ** p < 0.01; and *** p < 0.001. OLS — Ordinary Least Squares regression.
Contrary to H1a and H2a, we find no evidence to suggest that the EP uses less determinate language to overcome the challenge of finding a shared position in the face of internal divisions. The effect of Intra-EP conflict is negative, in line with H2a, but not significant in either model.
Considering the effect of inter-institutional (EP-CN) conflict, we find that contrary to H1b, the EP does not introduce more vaguely formulated amendments when its preferences do not align with those of its co-legislator. Instead, in line with H2b, the results show the EP-CN conflict has the opposite effect. On average across legislative provisions, parliamentary positions become more concrete relative to the Commission proposal as interinstitutional conflict increases: a one-unit increase in conflict is associated with a 0.11 percentage points decrease in vagueness. Panel (a) of Figure 2 illustrates this effect by plotting the predicted Increase in vagueness across levels of inter-institutional conflict. The figure shows that, on average, the lowest levels of conflict are associated with the use of less determinate language in the parliamentary position relative to the Commission’s proposal. Conversely, at higher levels of conflict (above 2.3 on the measurement scale of 0.01 to 2.53), the EP employs more precise language, thereby imposing stronger obligations and limiting implementation flexibility.
Effects on the change in vagueness between proposal and EP position.
This suggests that when the EP and the Council diverge, the EP has stronger incentives to lock its position in the adopted legislation and, thus, ensure the intended policy implementation (Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024) so that the law-as-implemented closely reflects the law-as-adopted. Reducing vagueness of the Commission’s proposals helps to achieve this goal by providing clear and firm instructions for law application and limiting discretion of implementing agents (Staton and Vanberg, Reference Staton and Vanberg2008; Williams, Reference Williams2018).
The results yield support for H3 on the influence of increased public opinion division on the EP’s use of vague language when amending the Commission’s proposals. The positive significant effect of this variable suggests that the EP considers public opinion and avoids presenting firm positions when EU citizens grow more divided in their policy preferences. The effect remains robust across both model specifications. Based on Model 2, on average, the hypothetical increase in public opinion division from its minimum (−1) to its maximum (1) between the times of Commission proposal and the first inter-institutional negotiations leads to about a 0.37 percentage point increase in parliamentary position vagueness relative to the text proposed by the Commission. Panel (b) of Figure 2 further shows that when opinion division decreases or remains unchanged between the proposal and the trialogue negotiations, it has no significant effect on EP position vagueness. In contrast, an increase in opinion division is associated with the use of less determinate language in the EP text.
This is a small but notable effect size given that the average increase in the vagueness of legislative proposals that the EP introduces with its amendments is 0.102 percentage points with a standard deviation of 0.381. Crucially, in the legislative texts, small lexical changes, captured by the application of obligation strength dictionaries, can represent substantive change in meaning, for instance, if policymakers introduce conditional language or use indeterminate words such as may or might instead of must. This substantially impacts law interpretation and application as well as the flexibility for future law adaptation in the face of new developments (Li, Reference Li2017; Williams, Reference Williams2018).
In line with previous studies (Sorace, Reference Sorace2022), this result suggests that the EP is responsive to public opinion when preparing its positions for legislative negotiations. Although this does not constitute policy responsiveness in its most common definition—policy change following accordingly public opinion change (Stimson et al., Reference Stimson, Mackuen and Erikson1995; Soroka and Wlezien, Reference Soroka and Wlezien2010)—increasing the vagueness of prospective laws amid divided public allows the EP to signal responsiveness to the citizens (Hagemann et al., Reference Hagemann, Hobolt and Wratil2017) or, at least, avoid taking a side and, thus, potential audience costs.
Our last hypothesis (H4) is also supported. Panel (c) of Figure 2 shows the predicted increase in vagueness over the levels of substantive change to EU authority that the EP introduces to the Commission’s proposals. The hypothesized U-shape is clearly visible. This result indicates that, as expected, the EP tends to formulate its policy positions vaguely when it seeks to substantively change the supranational authority, either to curb or to advance more integrationist measures.
Furthermore, our results indicate that the EP adopts more precise amendments to legislative proposals in policy areas of higher public salience. This suggests that the EP seeks to ensure uniform law interpretation and implementation when EU policy actions are more visible and important to EU citizens. However, this effect should be interpreted with caution, as it is statistically significant only at the 10% level in Model 2. None of the other control variables has a statistically significant effect.
Our main results remain stable after several robustness checks (see Appendix 9). First, we re-estimated the models using alternative measurement of the DV. Second, we replicated the analyses with the addition of policy area and EP term (or year) fixed effects. Third, we included controls for mobilization of business actors in a policy field and a year and public opinion division in the member state of the rapporteur. While the models that include year fixed effects and opinion division in the rapporteur’s country yield somewhat more modest results with respect to the effect of public constraints, other results are fairly similar to the analyses presented above.
7. Discussion and conclusion
Research shows that policymakers use vague and concrete language strategically in the legislative texts they draft. They do so to overcome disagreements within and across policymakers (Staton and Vanberg, Reference Staton and Vanberg2008; Hansen, Reference Hansen2016; Mousmouti, Reference Mousmouti2019) and facilitate effective policy change (Li, Reference Li2017; Williams, Reference Williams2018; Li, Reference Li2019), or to give discretion to policy implementers and avoid costs of potential non-compliance (Staton and Vanberg, Reference Staton and Vanberg2008). However, the use of vague language in legislation can have negative effects, too. By leaving a lot of room for interpretation, it can reduce policy effectiveness and the uniformity of law implementation and application (Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024). Despite these important implications, we still know little about when legislative institutions, in general, and the legislative branches of the EU, in particular, advance vague laws. This is particularly surprising given the increased scholarly attention to the linguistic characteristics of the laws (Li, Reference Li2017; Williams, Reference Williams2018; Osnabrügge and Vannoni, Reference Osnabrügge and Vannoni2024).
To address this gap, we offered first insights on the strategic use of vague language in the EP’s amendments during legislative negotiations in the EU as a means to deal with institutional conflict and politicization. Drawing on the studies of legislative negotiations, we analyzed whether intra- and inter-institutional conflicts drive the EP to take a vague stance when it expects to face opposition as a way to foster compromise or, conversely, whether they incentivize it to avoid vague language so as to lock its positions in adopted legislation and policy outcomes. Borrowing from the literature on responsiveness in EU policymaking and the use of vague language to bridge political divides in electoral contexts, we further analyzed how the EP acts in a politicized environment. Specifically, we assessed whether the EP mitigates potential audience costs by adopting vague amendments in response to increased public opinion divisions over EU policy action and when suggesting amendments that increase contentiousness of legislative proposals, i.e. those with implications for EU policy authority.
While we find no robust evidence that intra-EP divisions shape the use of vague and concrete language in EP amendments, conflict between the EU co-legislators appears to dis-incentivize vague position formulation by the EP. This suggests that the EP, rather than unconditionally seeking compromise, attempts to secure a narrower interpretation of the legislative content during implementation when faced with strong Council opposition and, thus, high risk of policy-as-implemented significantly deviating from the policy-as-adopted (Williams, Reference Williams2018; Byers, Reference Byers2021). We argue that this could be explained with the institutional structure of the EU, where the policy implementers, i.e. member state governments, are simultaneously co-legislators (in the EU Council). This adds to the theoretical discussion on the trade-off in using vague language between securing compromise and shaping adopted policy.
Furthermore, as expected, increased disunity in public views over EU policy action at the time of first trialogue legislative negotiations relative to the time of proposal drives the EP to use vaguer language and reduce the strength of obligations imposed by the legislative act. This allows the EP to safeguard its reputation as the representative of the EU citizens even when there is no clear majority opinion for it to represent. Similarly, we show that the EP uses vague language when its amendments imply a substantive change to EU authority. Given the increased politicization of European integration, this finding suggests a strategic use of vague language in law formulation by the EP to avoid public backlash against its position on issues of increased public contestation.
Taken together, our results contribute, first, to research on ambiguity and vagueness as a deliberate strategy in multilateral negotiations (Byers, Reference Byers2021) and in national policymaking (Williams, Reference Williams2018; Mousmouti, Reference Mousmouti2019). We add to this literature by focusing on the initial stages of the negotiation process, thereby fleshing out legislators’ motivations to use vague language before they are shaped by the negotiation process itself. Our analysis shows no evidence supporting the proposition that vague position formulation is used to reach agreement at the early negotiation stage. Instead, it supports the rationale that institutional conflict incentivizes legislators to take more concrete stances as a way of securing that their positions are incorporated in adopted legislation and, eventually, in policy outcomes by limited implementation discretion.
Second, we also add to the literature concerned with the EP role in policymaking. With the growing importance of informal negotiations as the main bargaining forum and policymaking stage in the EU (Brandsma, Reference Brandsma2015; Haag, Reference Haag2022), the results add important insights into the strategies the EP employs when preparing for these negotiations. We find that the EP uses more concrete language when the conflict with its co-legislator is high. It may do so to steer the legislative compromise toward its own position. As the EP’s legislative influence is weaker when its preferences do not align with those of the Council (Kreppel, Reference Kreppel2018; Haag, Reference Haag2022), adopting a firmer, maximalist position at the onset of negotiations (Bicchi, Reference Bicchi2026, pp. 8–9) may help it to widen bargaining space and secure a final legislative outcome that is closer to its preferred outcome. Furthermore, by using more determinate language in its amendments and strengthening legal obligations when the preferences of the co-legislators diverge, the EP may help prevent potential agency drift during the implementation stage. This points to a mechanism of post-legislative parliamentary control on the supranational level (Griglio, Reference Griglio2019, pp. 132–134), whereby the EP promotes the achievement of envisioned policy goals and the consistency of law application across EU member states. Assuming that the EP pursues its mandate of representing the interests of EU citizens, it can contribute to a better alignment of EU policies with public views and, thus, decrease the alleged democratic deficit in the EU (Follesdal and Hix, Reference Follesdal and Hix2006). However, this hinges on the extent to which the Parliament is able to assert concrete amendments into adopted legislation, which is an important venue for future research.
Third, by linking vague language in EU law formulation to the dynamics of public opinion and the contentiousness of supranational authority and policy, we offer a novel explanation for policymakers’ strategic use of vagueness when legislating as a means to limit polarization, audience costs, and public backlash. In EU research, this contributes to the growing literature on the responsiveness and legislative behavior of supranational institutions in the face of the politicization of EU authority (Wratil, Reference Wratil2018; Koop et al., Reference Koop, Reh and Bressanelli2020; Yordanova et al., Reference Yordanova, Ershova, Khokhlova, Obaid Ul-islam and Glavaš2025a, Reference Yordanova, Khokhlova, Ershova, Schmidt and Glavaš2025b). We add to this research by showing that, in the face of competing policy views of EU citizens or potentially contentious legislation, the EP gives the Council members a room to tailor law implementation to national contexts and meet demands of their member state citizens (Steunenberg, Reference Steunenberg2019). In other words, while enabling the EP to respond to divided public opinion across the EU, vague parliamentary positions could also enhance the output legitimacy of the EU by allowing for governments’ responsiveness to domestic publics during implementation.
Despite these contributions, this study leaves several questions unanswered. By examining the Parliament’s initial negotiation positions, we do not directly assess whether its strategic use of concrete or vague language contributes to its success in inter-institutional bargaining. As we suggest above, this is an important venue for future research. Furthermore, we do not engage with the role of unintended ambiguity (Ringe, Reference Ringe2022, p. 180). A certain degree of text vagueness might stem from the stylistics and language constraints of the actors defining the EP’s policy position. However, as Ringe (Reference Ringe2022) discusses, the EP’s professional lawyer-translators and large cross-language teams involved in position formulation make unintentional formulations of the EP positions due to policymakers’ language skills unlikely. Instead, unintentional vagueness is likely to increase during the secluded trialogue negotiations where support by lawyer-translators is more limited. Whether this is the case and how it impacts the text of the adopted legislation thus presents an interesting topic for future research.
Acknowledgements
We are grateful to Natália Kubalová, Sarah Knabl, Daniel J. Whitten, and Bence Hamrák for their invaluable help with the data collection and Saad Obaid Ul-Islam for the technical assistance. We would like to thank Nils Ringe, Bjørn Høyland, and Fabio Franchino as well as the participants of CES 2023, EPSA 2023, Norface Governance Final Conference, EUSA 2025, and the Lunch Seminar at the Institute of Political Science (Leiden University) for the comments and suggestions on previous versions of the paper.
Author contributions
Conceptualization: A.K., N.Y., and A.E. Methodology: A.K., N.Y., and A.E. Data curation: A.K. Data visualization: A.K. Analysis: A.K. Writing: A.K., N.Y., and A.E. Supervision: N.Y. Funding acquisition: N.Y.
Data availability statement
Replication data and code are available at the PSRM Dataverse.
Declaration of interests
None.
Ethics and consents
The research meets all ethical guidelines, including adherence to the legal requirements of the study country.
Funding statement
This work was conducted as part of the project EUINACTION Willingness and Capacity for EU Policy Change, which was financially supported by the NORFACE Joint Research Programme on Democratic Governance in a Turbulent Age and co-funded by DFG, ESRC, NWO, and the European Commission through Horizon 2020 under agreement (No. 822166).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2026.10123. To obtain replication material for this article https://url.uk.m.mimecastprotect.com/s/xJNKCRoYYFrKAV85F9fxf1IrJ-?domain=doi.org.
Open access
