Conversations in conservation: revealing and dealing with language differences in environmental conflicts


*Correspondence author. E-mail:;


  • 1Applied ecology aims to translate research into policy recommendations. However, conflicts frequently develop if these recommendations propose a contentious course of action. A first step towards addressing such conflicts is to attempt to understand the values underpinning stakeholder viewpoints.
  • 2We develop a computer-aided Content Analysis to analyse the language surrounding environmental conflicts for insights into stakeholder values. Using the conflict arising over proposals to cull hedgehogs Erinaceus europaeus on several Scottish islands, we show how different stakeholder groups frame the problem in different ways.
  • 3Stakeholder groups supporting different courses of action (culling vs. translocating hedgehogs) use different arguments, the former emphasizing conservation and biodiversity, the latter focusing on animal welfare. Our method results in a graphical representation of this failure to agree on a common way to frame the issue.
  • 4Including texts obtained from media sources illustrates how the media can exacerbate environmental conflicts through the issues they emphasize and the vocabulary they use.
  • 5Synthesis and applications. Our method provides a simple means to quantify levels of stakeholder disagreement concerning potentially contentious environmental issues. Our results provide a starting point for the development of a quantitative, graphical tool for managers, where repeated analysis will aid in monitoring and managing conflicts. In addition, we provide a clear example of the role of societal attitudes influencing the effective implementation of ecological advice, which should encourage ecologists to become more aware of the social environment into which policy recommendations are to be launched and to ensure that their advice does not ignore important stakeholder values.


Applied ecology is judged on management relevance as much as on scientific excellence (Flaspohler, Bub & Kaplin 2000; Ormerod et al. 2002). Success depends both on translating science into recommendations, and on the implementation of these recommendations into policy. Even when recommendations are clear (e.g. Jackson 2001; Ellis & Elphick 2007), their implementation may prove difficult: policy decisions about the conservation or exploitation of limited natural resources are beset by controversy (Ehrenfield 1991; Mills & Clark 2001; Peterson et al. 2002; Redpath et al. 2004; Perry & Perry 2008; Swart & van Andel 2008). This arises due to a lack of consensus regarding appropriate management goals between a range of stakeholders including lobby groups, managers and users of biodiversity as well as researchers and advisers from a broad range of natural and social disciplines (Peterson et al. 2002; Redpath et al. 2004).

When stakeholders differ fundamentally in philosophies, values and aspirations, damaging conflicts can arise (Peterson et al. 2002). Examples of costly value-driven conflicts include campaigns against plans to cull problem populations of charismatic vertebrates (Minteer & Collins 2005a; Ellis & Elphick 2007; Perry & Perry 2008) and arguments over the potential dangers of genetically modified organisms (Gray 2004). Given the costs in time, money and damaged public relations that such conflicts can accrue, it is clear that successful participatory approaches to natural resource management must seek to understand the different value systems of stakeholders (Endter-Wada et al. 1998). Such efforts will aid applied ecologists to become more engaged with the ethical and political implications of their research (e.g. Minteer & Collins 2005a,b), and to deal with increasing pressure to engage actively in decision-making (Mills & Clark 2001) through a better understanding of the social environment into which management recommendations will be launched.

There is a long history of analyses of the power of language in the social sciences (Diefenbach 2001), but its role in environmental debates is often not appreciated by ecologists. Stakeholders frame contentious issues in different ways (Entman 1993; Thompson & Lapointe 1995; Miller & Riechert 2001), and ecological phrasing such as the ‘militaristic’ language often used in discussions of invasive species may result in misunderstanding and a loss of scientific credibility among sections of the public (Larson 2005). Equally problematic, even those with similar scientific training can interpret specific terms in different ways (Christie et al. 2006; Holt 2006). Such issues can be identified and clarified through analysis of the language used in an environmental debate (Miller & Riechert 2001), but to date, most analyses in an environmental context have concerned temporal trends in the frequency with which topics are reported in the media (Bengston & Fan 1999; Bengston 2000; Bengston, Xu & Fan 2001). We argue that appropriate, quantitatively rigorous techniques exist to allow more fundamental analyses of language aimed at identifying the drivers of environmental conflicts.

In order to asses whether such analyses can help us to understand, and potentially to mediate, value-driven conflicts over biodiversity issues, we build on the strong theoretical foundation of Content Analysis (Holsti 1969; Weber 1990; Hodson 1999; Diefenbach 2001) to develop a flexible, computerized tool to assess features of stakeholder language use. We demonstrate the utility of this method through a detailed analysis of a representative conflict concerning plans to cull hedgehogs to protect bird populations on Scottish islands. In addition, we emphasize that language is both important and amenable to simple analytical techniques which will help ecologists to develop an awareness of the social environment surrounding an environmental issue, to search for common ground between stakeholders, and to frame management recommendations in order to reduce the likelihood of conflict.

The biodiversity conflict

The islands of Benbecula and North and South Uist in the Outer Hebrides, Scotland support internationally important breeding populations of ground-nesting shorebirds (Charadrii) (Jackson & Green 2000). Hedgehogs Erinaceus europaeus L. were introduced from the mainland to South Uist in 1974 in an attempt to control garden pests; since then, the species has spread to all three islands. Hedgehogs are strongly implicated in the subsequent significant declines of several species of shorebird (Jackson & Green 2000; Jackson 2001), leading Scottish Natural Heritage (SNH) to propose a cull (Uist Wader Project 2002). This was vehemently opposed by animal welfare and hedgehog conservation groups, who promoted the translocation of hedgehogs to the mainland through ‘hedgehog rescue’ operations which offered bounties of up to £20 for each hedgehog brought to them alive. The conflict escalated as the local and national media picked up on this ‘prickly issue’, and interest reignited with each annual announcement of the cull.

In common with many biodiversity conflicts, the main point of contention is not directly financial (although economic arguments have been made on both sides). Rather, both sides value ‘nature’ over at least some economic considerations (Perry & Perry 2008) but differ in the focus of their ethical concern (Ehrenfield 1991), from valuing above all the lives of individual animals (hedgehogs in this case) to cherishing instead the integrity of populations, species and ecosystems (shorebird assemblages). Other issues common to many conflicts include the highly charged language concerning ‘alien’ species (Larson 2005), debates about the role of ‘technical elites’ (scientists) in determining environmental policy (Mills & Clark 2001; Lach et al. 2003), accountability and hidden agendas of government agencies (Thompson & Lapointe 1995), and the influence of media coverage on the trajectory of environmental conflicts (Bengston & Fan 1999; Bengston 2000; Bengston et al. 2001).


texts produced by stakeholders

Thirty nine stakeholder texts, representing 13 organizations or individuals, were identified from the results of a search using the term {hedgehog* AND uist*} in Google ( Texts were analysed in their entirety except for the few where the issue constituted a small part of a longer document, in which case we analysed relevant sections only. Texts ranged from 139 to 4062 words (median ± m.a.d. = 538 ± 363·2 words, total = 28 602 words), and were produced between January 2002 and September 2005, except for one from January 2001. Based on the overall tone and content of stakeholder websites, each organization was classified as either pro-hedgehog (H, N = 23) or pro-wader (W, N = 16). In this conflict, such assignations are not contentious.

texts from the media

We obtained relevant news stories and opinion pieces (excluding letters to the editor) from print magazines, online news sources, and broadcast media identified in the Google search described above, and searched UK newspapers using the term {(hedgehog* OR wader*) AND (uist*) AND (cull* OR threat*)} in NewsBank ( We retained 466 media texts from 47 publications (18 to 2359 words per text; median ± m.a.d. = 230 ± 203·9 words; total = 135 492 words), published between February 2002 and December 2005 apart from single texts from March 1998, January 2000 and March 2001. Broadsheet and tabloid newspapers were represented, as well as online, magazine and broadcast media, but initial analyses revealed few differences between media categories in their coverage of this conflict and thus we considered all media articles as a single group here. We used publication as the unit of analysis, taking mean values from all articles appearing within each publication, although our conclusions are unaffected if individual articles are analysed.

quantitative analysis of texts

We used a Content Analysis (CA) approach to analyse the texts. A primary purpose of CA is to reveal the issues important to different groups (Weber 1990), and it has the advantage of transforming texts into data amenable to quantitative analysis (Hodson 1999). CA involves defining a series of vocabulary categories to define an issue, and then classifying the vocabulary within the sample of texts into these different categories (Hodson 1999). We defined vocabulary categories based on our knowledge of the conflict and refined them through initial classification of samples of vocabulary into the list shown in Table 1. We also calculated three summary statistics for each text (Table 1). We used a list of all words occurring across all texts and placed each word into an appropriate category (or left it uncategorized). Categorization is thus separated from the texts, allowing greater objectivity and eliminating the possibility of skewing the categorization in favour of particular theories about the different stakeholders (Hodson 1999). One thousand six hundred and forty-three (47%) of 3462 words in the stakeholder texts and 3454 (47%) of 7411 words in media texts were classified into a category. Although classifications were performed by one coder (T.J.W.) to ensure internal consistency, high agreement (> 70%) with categorization of 20% of the words by an independent coder (familiar with the methods but not the conflict) show that our results are robust to coder identity. The context of all categorized words was checked by scanning the sentences in which they appeared for features such as negative qualifiers likely to change the meaning of a word. Finally, the proportion of words in each text falling into the 15 categories was calculated. Generally, < 50% of words in a text were classified, with a maximum of 12% of the words in a text being classified into any one category; thus, the proportions are not constrained by the values for other categories. The advantages of computerized CA are well documented (Bengston 2000; West 2001a,b) and for all text manipulation, extraction of vocabulary lists and processing of categorized vocabulary, we used atat: A Text Analysis Tool (Webb 2006; available from the authors on request), a Visual Basic for Applications (VBA) procedure designed specifically to use commonly used existing software to perform quantitative analysis of texts.

Table 1.  The categories used to classify the vocabulary in the texts. Specific vocabulary refers to a specific action or group of objects; Legitimacy refers to means of gaining legitimacy either by referring to organizations by name, or by using scientific terms and titles. Style of language identifies conciliatory, antagonistic and other value-laden terms, as well as use of informal terms (e.g. personal pronouns) and slang. Finally, some simple summary data were extracted from the texts
  • *

    Refers to Mrs Tiggywinkle, the hedgehog heroine of a popular children's book by Beatrix Potter.

  • Terms of interest consisting of more than one word were identified prior to analysis, and compressed into a single word in the texts so that incorrect classification could be avoided. For instance, we did not want to classify separately the three words in Scottish Natural Heritage.

  • GBP, Great British Pounds (sterling).

  • §

    The number of quotes in a text was estimated as the number of closing quotation marks (’).

Specific vocabulary
 Killingcull, kill, shoot, slaughter
 Welfare/caring/translocationcare, cruel, humane, rescue
 Hedgehogshedgehog, Erinaceus, Tiggywinkle*
 Birdsdunlin, ground nesting, ornithologist, wader
 Wildlife/conservation/biodiversityamphibian, conservation, environmentalist, habitat, sand dunes
 Agriculture/land usecroftland, farming, forestry, pasture
 The HebridesBenbecula, Hebridean, Isles, Uist
 Economic/employmentbuy, cash, cost, expensive, GBP
 Legal/statutory/politicalban, Birds Directive, lawful, policy, rule
 Science/technologyassess, biological, data, ecologist, professor
 Specific organizationsAdvocates for Animals, Scottish Natural Heritage, RSPB
Style of language
 Conciliatorycoalition, compromise, goodwill, optimistic, sensible, welcomed
 Antagonisticaccuse, confront, devastate, flawed, lies, rejecting, unfounded
 Other value judgementsbeautiful, complicated, dignified, glad, opinion, reluctant, weird
 Informal/slangain't, cuddly, I, my, OK, squelching, veggie
 Structure of text 
 Log(number of words) 
 Mean words.sentence−1 
 Mean quotes.paragraph−1§ 

The 18 variables in Table 1 were reduced using Principal Components Analysis (PCA), by extracting Principal Components (PCs) using a singular value decomposition of the standardized data matrix. We examined divergence in the PC scores between the groups (stakeholders: H, W; media: M), defined above, using both multivariate (manova) and univariate (anova) approaches, and we used a linear Discriminant Function Analysis (DFA), employing leave-one-out cross-validation to test the accuracy with which a text could be placed into the correct stakeholder group when these groups were defined using all other texts. We used univariate comparisons between the groups separately for each category to identify where differences between groups lie. We analysed stakeholder texts, and then stakeholder and media texts combined. All analyses employed r 2·4·1 (r Development Core Team 2006).

To illustrate language differences within vocabulary categories, we examined ‘Key Words in Context’ (KWIC; Weber 1990) using atat to list all sentences containing ‘economic’ vocabulary. We present a qualitative comparison of economic vocabulary use between pro-hedgehog and pro-wader groups. Next, we scored each of the 59 words in the ‘killing’ category on a scale of –2 (highly emotive: emphasizing the brutal, unpleasant aspects of killing) through 0 (neutral: free of emotive baggage) to +2 (highly euphemistic: avoidance of upsetting vocabulary). Ten coders independently scored all 59 words, mean scores were taken for each word and compared between stakeholders and publications. Finally, we analysed the relative frequencies and positions within texts of specific terms: the names of the principal SNH and Advocates for Animals spokesmen, the organizations ‘Uist Hedgehog Rescue’ and ‘Uist Wader Project’, and the position within a text of the first mention of hedgehogs compared to the first mention of birds.


stakeholder texts

Thirty per cent of the variation in language between texts is explained by the first PC, 73% by the first six PCs. Language defined by these six PCs differs significantly between H and W stakeholders (manova, Pillai's trace = 0·71, F6,32 = 12·91, P < 0·00001), largely due to differences in PC1 (Fig. 1; difference between H and W texts in PC1: R2 = 0·67, F1,37 = 75·54, P < 0·00001). The loadings on PC1 (Table 2) show that pro-hedgehog texts (low values of PC1) focused on killing, economics, hedgehogs, welfare, and emotive or informal vocabulary, whereas pro-wader texts (high values of PC1) used scientific language and vocabulary concerning birds, agriculture, wildlife and the Hebrides. anova tests on individual categories support these differences (Fig. 2). The DFA showed that, using classifications based on the remaining texts, 34 of 39 texts (87·2%) were correctly assigned to H or W.

Figure 1.

The first two Principal Components for 23 pro-hedgehog texts (H), 16 pro-wader texts (W), and 47 media texts (M). Minimum Convex Hull polygons are fitted around the respective groups of texts. The right hand panels show differences between the stakeholder and media groups on PC1 and PC2, with different shading indicating significant differences between the groups. See Table 2 and text for details of the loadings on each factor.

Table 2.  The loadings of the variables described in Table 1 on the first two principal components, for stakeholder texts only, and for stakeholder and media texts combined
 Stakeholder textsCombined texts
Figure 2.

The proportion of words in each text falling into representative categories (first eight plots), and mean text length (log-transformed) for each groups of texts. Different shading within a plot indicates significant differences at the 5% level from a univariate linear model on that variable alone. We have not corrected significance for multiple tests, as we intend the plots to be illustrative, in support of the conclusions from the multivariate analyses discussed in the text.

stakeholder and media texts

Twenty-two per cent of the variation between texts is explained by PC1, and 69% by the first six PCs, with significant differences in language use revealed between H, W, and Media (M) texts (manova, Pillai's trace = 1·06; F12,158 = 14·90, P < 0·00001). Each group differed significantly from the others. PC1 separates W from H and M texts (Fig. 1; anova, R2 = 0·52, F2,83 = 47·04, P < 0·00001). High values of PC1 again reveal the scientific language and vocabulary concerning birds, wildlife and agriculture employed in W texts, as opposed to the informal, antagonistic or conciliatory language relating to killing, economics, animal welfare and hedgehogs found in H and M texts (low values of PC1; Table 2, Fig. 2). H texts (high values) were separated from W and M texts (low values) on PC2 (Table 2, Fig. 1).

differences within vocabulary categories

Qualitative differences exist between the stakeholders in their use of economic language. Advocates for Animals (H) stressed the expense of the cull: ‘So far, SNH's killing policy has cost the taxpayer hundreds of thousands of pounds’; ‘Is this really how the public expects their taxes to be spent by SNH, a government-funded so-called “conservation” organisation?’. Pro-wader groups either emphasized the economic importance of shorebirds (‘The maintenance of machair biodiversity is vital to the Uist economy’, Uist Wader Project) or simply stated costs (‘The cost of the Spring programme is approximately GBP 62 000’, SNH).

Pro-hedgehog groups used more emotive language to describe killing than pro-wader groups (mean H = –0·26 ± 0·045; mean W = –0·05 ± 0·075; F1,32 = 7·82, P = 0·0087; Fig. 3); media texts had similar scores to the hedgehog stakeholders.

Figure 3.

Differences between pro-hedgehog (H), pro-wader (W) and media (M) texts in the use of different kinds of vocabulary concerning killing; different shading indicates differences significant at the 5% level. The width of each bar is proportional to the sample size within each category. Positive scores indicate euphemistic vocabulary, negative scores are emotive vocabulary.

Spokesmen for SNH and Advocates for Animals were named by a similar proportion of the 466 news stories (18% and 21%, respectively; binomial test for difference between proportions, P = 0·9241), whereas significantly more stories (36%) named the Uist Hedgehog Rescue organization than the Uist Wader Project (11%; P = 0·0459). Although most media texts (88% of stories, 96% of publications) mentioned both birds and hedgehogs, hedgehogs were usually mentioned within the first 5% of a story (mean ± SE = 4·4 ± 0·60%), birds usually considerably later (42·7 ± 3·68%, non-occurrences scored as 100%; paired t-test with unequal variances, t = 11·45, d.f. = 46, P < 0·00001).


Our analyses have demonstrated that Content Analysis can help us to understand value-driven conflicts over biodiversity issues by clarifying differences between stakeholder groups. The debate over the removal of hedgehogs from the Outer Hebrides is clearly framed differently by pro-wader and pro-hedgehog stakeholders (a story of broader biodiversity and land management conflicting with an iconic mammal, cruelty and animal welfare). It is clear that basic agreement on what exactly are the defining issues, a critical first step in resolving environmental conflicts (Peterson et al. 2002), has yet to be reached. Media coverage has played a significant and not always helpful role in perpetuating and polarizing this conflict, and in general, the language of the print media mirrored that of the pro-hedgehog lobby.

Whilst the present analysis is squarely focused on a very local issue (albeit one typical of a range of conflicts between animal welfare campaigners and environmental managers; Perry & Perry 2008), we foresee many generic applications of the basic approach adopted here. First, it provides a simple, quantitative snapshot of the social environment surrounding a conflict. Thus, in our case study, a mediator could formally demonstrate to stakeholders the lack of common ground, providing an authoritative basis for advocating that the pro-wader group should produce a statement on animal welfare, and the pro-hedgehog group should produce one on bird conservation. This would provide a clearer demarcation of the differences to be overcome. Although such snapshots can be useful, applying the methods in an iterative fashion is likely to prove even more valuable. CA of written texts is particularly suited by its cumulative nature to later additions and extensions (Hodson 1999). Thus, the relative stances of different stakeholders could be compared repeatedly as a conflict progresses to determine whether groups are diverging or converging. Such an approach would be particularly useful in assessing the effects on stakeholder attitudes of particular policy statements or management interventions, and develops previous efforts to use language to monitor society's attitudes towards environmental issues (Bengston & Fan 1999; Bengston et al. 2001).

The approach depends on the availability of sufficient text to analyse, and is thus most appropriate for the analysis of established, ongoing conflicts, rather than in highlighting potential conflict. However, large-scale public consultation exercises, for example the GM Nation debate in the UK (DTI 2003) over the release of genetically modified organisms, rapidly produce large volumes of text. In general, words are not in short supply when issues are contentious. Applied to such exercises, our methods could use PCA or clustering methods to identify groups of stakeholders related by language use, rather than the a priori grouping we employed (confirmed by DFA). Groups based on specific vocabulary will identify stakeholders with common perspectives, whereas groups based on tone (e.g. ‘antagonistic’ vs. ‘conciliatory’) will provide information concerning those stakeholders likely to be most amenable to compromise, or those most likely to require more concerted negotiation.

Context is clearly important to interpretation of results: the analysis may not identify distinctions between stakeholders, yet if all are using highly antagonistic vocabulary, conflict may be intense. We have shown how considerations of language use within vocabulary categories can be useful in this respect, for instance, the different uses of economics vocabulary by different stakeholder groups, demonstrating how CA can combine the respective strengths of quantitative and qualitative analyses (Hodson 1999). Subtle biases in media coverage were also revealed, for instance, the tendency for hedgehogs to be mentioned earlier in an article than birds; similar methods could prove useful in determining the responses of different stakeholders to an important scientific report or piece of legislation. Other questions that could be addressed include using transcripts of spoken language to compare the ‘official line’ of an organization (determined from press releases) with opinions expressed by individuals during interviews; likewise, opinions of local campaigners may diverge from the official policy expressed by their organization at the national level (Thompson & Lapointe 1995). Investigating ecological correlates of conflict likelihood or intensity would also be possible. For example, the cull of American mink Mustela vison on the Uists has been far less controversial (21 stories identified from NewsBank) than that of the UK native hedgehogs (398 stories); there may be general differences between native and exotic species, or between popular birds and mammals compared to reptiles or invertebrates (Perry & Perry 2008).


Applied ecologists rightly consider the production of policy-relevant management recommendations as a key research output (Flaspohler et al. 2000; Mills & Clark 2001; Ormerod et al. 2002), but these can prove contentious, resulting in damaging environmental conflicts. Dismissing public concerns as unscientific or misled is unhelpful (Leach 2007), but there is little recognition in the ecological literature that the successful implementation of management recommendations rests on public acceptance (Endter-Wada et al. 1998; Vining, Tyler & Kweon 2000; Swart & van Andel 2008). Even when ethical concerns are considered (Baker et al. 2005; Ellis & Elphick 2007), steps to minimize conflict may be insufficient to overcome the objections of a small number of concerned and emotional citizens (Vining et al. 2000). Of course, tough decisions still have to be made, but we suggest that an explicit consideration of stakeholder values and a search for common ground can lead to a more systematic discussion of the costs and benefits (including public acceptability) of different management options. Short delays in policy implementation caused by taking these steps are likely to be more than offset by the avoidance of legal costs and long-term public relations disasters (Perry & Perry 2008). Critical to such efforts is the description of stakeholder values, and the methods introduced demonstrate that this is possible through a simple quantitative analysis of language.


Thanks to members of the UKPopNet project Framework for sustainable livelihoods, biodiversity change & conflict resolution for many helpful discussions and suggestions, to Rachael Maskill for help with vocabulary coding and testing the software, and to Rob Freckleton for comments. Three anonymous reviewers provided useful comments on a previous version of this manuscript. This work was funded by NERC through UKPopNet project 3b.