Conservationists’ perspectives on poverty: an empirical study

1. Biodiversity conservation interventions have long confronted challenges of human poverty. The ethical foundations of international conservation, including conser -vation's relationship with poverty, are currently being interrogated in animated debates about the future of conservation. However, while some commentary ex ists, empirical analysis of conservation practitioner perspectives on poverty, and their ethical justification, has been lacking thus far. 2. We used Q methodology complemented


| INTRODUC TI ON
The ethical foundations of conservation are under scrutiny. This is evident in the animated so-called 'new conservation' (Holmes, Sandbrook, & Fisher, 2017;Kareiva & Marvier, 2012;Soule, 2013), and the Half Earth/Whole Earth debates (Büscher et al., 2017;Wilson, 2016). This paper interrogates contemporary debates about one of the most ethically urgent issues that intersect with conservation: human well-being and its converse, poverty Howe, Corbera, Vira, Brockington, & Adams, 2018;Lehmann, Martin, & Fisher, 2018). Conservation has long confronted poverty because of the spatial intersection at a global scale of biodiversity and of human development challenges, noted in Fisher and Christopher (2007), and described by E. O. Wilson as an 'awful symmetry' (Wilson, 1992, p. 260). Until relatively recently, mainstream sustainable development thinking considered poverty as a chief cause of environmental degradation, and hence appropriate target of conservation (Duraiappah, 1998). However, social scientists have also documented how conservation benefits tend to accrue to the global community, while the disbenefits are localized and can exacerbate poverty, through displacement or restricted access to natural resources associated with protected areas (e.g. Brockington & Igoe, 2006). Brockington (2009) therefore argues that we should see conservation processes as spreading fortune and, crucially, misfortune around the landscape. The global political economy of conservation makes social impacts particularly contested; the conservation movement had a colonial genesis (Adams, 2004;Grove, 1995), and there are continuing legitimacy questions raised by the influence of conservation organizations headquartered in wealthy countries with the power to shape the relations between society and nature in poorer places (Chapin, 2004;MacDonald, 2008). Hence, the conservation/poverty nexus forms an arena of competing imperatives and obligations towards human and non-human nature, and the resulting trade-offs are among the most dramatic, or 'tragic' (Martin, 2017), and therefore contested and debated, within the broader field of environment and development.
The 1987 World Commission on Environment and Development (WCED, or Brundtland report) had a wide remit, but was particularly influential in framing linkages between poverty and conservation. Indeed, it is considered so influential that Duraiappah (1998) described it as a 'blueprint', and it dominated how international environment and development fora considered poverty at least through the 1990s. Poverty was characterized as a fundamental threat to sustainability and a central driver of resource degradation; 'poor and hungry' people were portrayed as driven by survival to overexploit and degrade resources (WCED, 1987, p. 28). The logic presented by Brundtland and institutionalized at the 1992 Earth Summit was that if poverty was to be addressed, instrumental benefits could accrue for conservation.
Developments in the policy realm continued to shape these de-  (Roe, 2008).
More concretely, the Durban Action Plan developed targets for protected areas to strive to alleviate poverty and in no case exacerbate poverty, and committed that all existing and future protected areas be required to comply with the rights of indigenous and mobile peoples and local communities (IUCN, 2003, in Roe, 2008. Concurrently, in an article taken very seriously within the movement, Chapin (2004) attacked the social impacts and human rights records of a number of the prominent, western headquartered conservation NGOs. These developments in the 2000s led to a moment of reckoning regarding conservation's social impacts and the exacerbation of poverty.
The establishment of the Conservation Initiative on Human Rights (CIHR: http://www.theci hr.org/), a network of organizations integrating rights-based approaches in conservation, was in many ways attributable to these developments through the 2000s, taking impetus from the Durban Congress (CIHR, 2014).
Hence, the instrumental logic of the Brundtland report yielded to a more normative logic for conservation to take account of poverty through the 2000s. In contrast also to the Brundtland particularly illuminates shifts in thinking in the 30 or so years since the Brundtland report, and reflecting new global realities.
6. This analysis serves to update, parse and clarify differing perspectives on poverty within the conservation, and broader environmental movement, to illuminate consensual aspects between perspectives, and reveal where critical differences remain.

K E Y W O R D S
conservation organizations, discourse analysis, poverty, Q methodology logic which conceptualized strong alignment between action on conservation and poverty alleviation, there developed increasing recognition of trade-offs and hard choices between these goals (McShane et al., 2010). The upshot of all of these developments, however, was that it became the norm for conservation organizations to seek to engage with matters associated with poverty for both ethical and pragmatic reasons (Walpole & Wilder, 2008).
Without necessarily denying the linkages drawn above, a related but distinct debate continued during the 2000s as to whether it was appropriate or effective for conservationists to attempt to address poverty issues, and some argued that this might detract from organizational missions (e.g. Sanderson & Redford, 2003), and core conservation activities (Terborgh, 1999). In this perspective, poverty and conservation are characterized as separate goals (Robinson, 2004). This mirrors the Tinbergen principle in economics which advocates one policy instrument per policy target (Klein, 2004). Redford, Levy, Sanderson, and de Sherbinin (2008) argue that there is actually very little spatial coincidence of global poverty hotspots with biodiversity in 'wild' areas (as a result of low density of poverty), and thus a genuine focus on addressing poverty would detract substantially from the core mission of conservation.
Effectively capturing a number of the positions described above, Adams et al. (2004) develop an influential typology of the different perspectives on relationships between poverty and conservation, which forms a central reference point for this paper. It is worthy of note, however, that this typology was derived from literature and deep knowledge of the field, rather than with reference to primary empirical material: 1. The first position considers that 'poverty and conservation are separate policy realms' (p. 1147).
2. The second position considers that 'poverty is a critical constraint on conservation ' (p. 1147). This is the Brundtland position described above, effectively an instrumental argument.
3. The third position takes a normative stance that 'conservation should not compromise poverty reduction ' (p. 1147). This is a manifestation of the 'do no harm' principle.
4. The fourth position considers that 'poverty reduction depends on living resource conservation' (p. 1148); in other words, this reverses the causal Brundtland logic and postulates that environmental degradation results in poverty; therefore, conservation can be promoted on the basis of supporting livelihoods. This position would tend to prioritize harvestable resources above species-or biodiversity-conservation Howe et al., 2018). There are parallels between this position and the 'environmentalism of the poor', ideas associated with Guha and Martínez-Alier (1997;Martinez-Alier, 2002) that the poor have a very considerable stake in the responsible management of the environment, for livelihood considerations.
The transitions in thinking we have traced through the literature constitute an anthropocentric turn seen more broadly in environmental management imperatives, but especially noticeable in conservation because of a stronger tradition of ecocentric protectionism. Conservation is increasingly justified through peoplecentred rationales, and attempted through people-centred approaches (Mace, 2014). The contemporary 'new conservation' debates turn on a hinge of whether conservation should be anthropocentric (Holmes et al., 2017), and Sandbrook, Fisher, Holmes, Luque-Lora, and Keane (2019) find that 94.7% of a global sample of 9,264 conservationists are in favour of people-centred conservation. The mission and public policy statements of most international conservation organizations show increasing attention to local livelihood issues, indigenous rights and poverty (Roe, 2008). Much organizational literature promotes the idea that human well-being depends centrally on ecological health (e.g. Conservation International, 2015;WWF, 2018).
Yet, there remain enduring concerns about instances where human rights or well-being have been compromised by conservation (e.g. Brockington & Igoe, 2006;Dowie, 2011;Reuters, 2019;Survival International, 2019). Hence, there remain questions about how far the anthropocentric turn goes, particularly beyond rhetoric to implementation. Ongoing concerns about the social impacts of conservation suggest that genuine commitments to poverty alleviation are more demanding, requiring more effort to achieve, than a generalized orientation towards people that seems to characterize the communications and direction of travel of many environmental organizations.
The aim of this paper is to investigate the contemporary discourses among conservation practitioners on poverty and conservation. The novel contribution this paper offers is an empirical analysis; existing commentary and analysis of historical developments are described above, but systematic and contemporary empirical analysis of practitioner perspectives is lacking.
In particular, there is a need to examine how far commitments go towards poverty alleviation, beyond a widely noted generalized anthropocentrism. We provide an in-depth analysis of how debates about poverty and conservation are manifest in contemporary conservation organizations, examining organizations headquartered in the global north, but also compare and contrast these with perspectives from divergent illustrative contexts in the global south. We employ a novel combination of Q methodology and qualitative analysis in a discourse analytical approach to scrutinize the ethical commitments of conservationists in a context in which the ethics of conservation, and particularly its social implications, are intensely debated (Holmes et al., 2017;Sandbrook et al., 2019). More generally, Sandbrook, Scales, Vira, and Adams (2010) note a lack of research on conservation values and ethics, which we address empirically in this paper. We use the Adams et al. (2004) framework outlined above to interrogate the discourses we elicit and to reflect on their contemporary implications. In line with Adams et al. (2004), and Howe et al. (2018), we offer this effort to parse and clarify differing perspectives on poverty within the conservation movement to illuminate consensual aspects between perspectives, and show where critical differences remain.

| ME THODOLOGY AND ME THODS
We employed a discourse-analytical approach to understand the perspectives of conservation professionals, drawing upon triangulated data from secondary materials, qualitative key informant interviews and Q methodology. Q methodology enables the comparison of individuals' ranking of statements to explore the structure and form behind subjective positions, giving the means to combine the qualitative study of perceptions with the statistical rigour of quantitative techniques (McKeown & Thomas, 1998).
Respondents consider statements reflecting various perspectives on a topic, and place them on a grid conveying agreement and disagreement (see Figure 1). Q supports the analysis of how subjective positions are shared by people, rather than with their prevalence in a population, the domain of conventional surveys. Q studies intensively analyse relatively small populations, often purposively sampled, rather than aiming to be statistically representative of larger populations.
Q methodology is an increasingly popular tool in the social science of conservation (Fisher & Brown, 2014;Sandbrook, Fisher, & Vira, 2013;Zabala, Sandbrook, & Mukherjee, 2018). While Q is now relatively established in conservation science (Zabala et al., 2018), it has not been used previously to investigate perspectives of conservationists on poverty. It is a powerful tool for identifying and analysing discourses and provides particularly strong triangulation in conjunction with more conventional qualitative analysis. In particular, initial impressions of qualitative data can be systematically investigated with Q methodology (Fisher & Brown, 2014;Zabala et al., 2018), and qualitative data can support the interpretation of discourses elicited with Q methodology (Watts & Stenner, 2012).
This study employs Q methodology in this triangulation role, accompanied by qualitative data. This combination allows us to investigate systematically the perspectives of conservation practitioners on the ethics underpinning and motivating their activities related to local communities and poverty.
Respondents for interviews and Q methodology were selected purposively to represent the broadest range of perspectives from international and national contexts on the issues of conservation and poverty in the global south. The total number of respondents engaged for this work is 39. An 'international' dataset comprised 14 respondents who work for organizations headquartered in North America and Europe, with at least some initiatives in the global south. This included all mainstream, prominent, conservation organizations, and further, smaller organizations representing a diversity of approaches, for instance, focused particularly on charismatic or endangered species, or conservation with development. Some groups were members of the CIHR. Because of this dense sampling (see Table 1) of large, mainstream international conservation organizations, we make representative claims about this group from our findings.
The sampling strategy also sought to include the perspectives of a number of respondents from national-level conservation organizations, to investigate aspects of debates about conservation and poverty in national settings. These country case studies were selected to be illustrative of widely differing geographies in Africa, Asia and Latin America and the countries chosen were Bolivia, China, Nepal and Uganda. The country cases were chosen partly for familiarity to the research team, allowing us to leverage deeper understanding from the findings. As case studies, these were not selected to be representative (Flyvbjerg, 2006), for instance at a continental scale, but instead to illustrate the character of debates manifest at national scale in diverse countries. Because of dense sampling (see Table 1), we can make claims about mainstream conservation within each national context, but we cannot generalize to any larger geographical unit. The combination of international and national organizations meant that some organizations were represented within both the international and national samples (Table 1).
A standardized semi-structured interview schedule and a structured Q methodology protocol was used throughout the research to ensure internal validity. Within organizations, we typically interviewed the respondent with the role most closely associated with local people. Respondents were asked to represent their own views, rather than to represent an organization, to avoid ambiguity and because of respondents' understandable reluctance to claim to represent sometimes large and complex organizations. Each interview has a unique code for reference. The sample is characterized in Table 1.
The sample contains 26 male and 13 female respondents. Data were collected over the course of a year, starting in April 2016. The research received clearance from the GeoSciences research ethics committee, University of Edinburgh. All respondents gave informed consent to participate in the study, and we maintain anonymity of respondents throughout.

F I G U R E 1
The Q methodology grid used for this study. Respondents were asked to allocate statements to cells reflecting their relative agreement with each statement To maintain complementarity between the qualitative analysis and Q, we sought a Q response from all interview respondents in the international and national samples. We have Q responses from 37 of 39 responses; there was attrition of two respondents (within the international sample) because they were unavailable at the time of the Q methodology interaction*. Table 1 describes the sample. A Q study starts by defining statements reflecting the range of perspectives on a topic and in relation to the research questions.
Stephenson (1952, p. 223) argued that the Q set (of statements chosen for a study) should be designed 'to suit the particular requirements of an investigation', and Watts and Stenner (2012) note that a balanced Q set is representative without core ideas missing. We used statements from primary empirical material (international sample interviews) and from a document analysis of secondary material released by sampled conservation organizations, which we had previously coded 1 for qualitative analysis. We also incorporated statements that we developed to ascertain responses to specific ethical principles. This combination of primary and secondary material is acceptable within Q methodology (Sandbrook et al., 2010;Watts & Stenner, 2012).
Adhering to the aims of representativeness and balance in statement design, we selected 32 statements from an initial list of 126, reducing the number by eliminating statements of lower relevance, or redundant statements whose meaning was more effectively conveyed by retained statements. Some statements were altered slightly for clarity or to reverse their meaning, to improve balance (Watts & Stenner, 2012). This Q set was then piloted with eight respondents, after which some small changes for clarity were made. The grid used is displayed in Figure 1, and respondents were asked to sort statements from 'most like I think' to 'least like I think'.
The internet software, htmlQ 2 was used to administer the Q survey with international respondents, who engaged with Q methodology some months after their interview. This software offered the submission of qualitative commentary on the statement rankings, which is important for interpreting Q results. Respondents in Bolivia, China, Nepal and Uganda were engaged with a paper version of the same Q exercise immediately after their interview. As regards the 'forced versus free' distinction, we encouraged respondents to follow the grid as closely as possible. Rather than as a requirement of statistical analysis, this encourages respondents to prioritize statements and place those most salient to them at the extremes (McKeown & Thomas, 1998;Watts & Stenner, 2012).
The interviews were conducted in English with international, Ugandan and Nepalese respondents. However, the Bolivian and Chinese respondents were engaged in Spanish and Mandarin, respectively. For the Bolivian and Chinese studies, Q statements were available in these languages as well as English (printed on the reverse of the cards), to aid understanding. Statement translations were undertaken by the bilingual leaders of country case studies, and checked using back translation to ensure validity. Where necessary, interview transcripts were also translated into English for qualitative analysis. Qualitative thematic analysis of interviews was undertaken using Nvivo software and these themes were drawn upon to illustrate results.
Q sorts were analysed using PQMethod software. Q methodology analysis focuses around factors, which are common orderings of statements (see Figure 2). We used a centroid factor analysis, in keeping with the principles of Q methodology (Watts & Stenner, 2012). Note: Numbering bears no resemblance to nomenclature of interviews in results section (to protect anonymity), but is included to demonstrate the size of samples.

TA B L E 1 Composition of sample of respondents
*An asterisk marks a respondent for whom there is an interview, but not a Q sort.
Three factors were extracted 3 and rotated. 4 Six respondents did not associate sufficiently with any one factor. Once factors are identified, the analysis becomes more interpretive and factors can be understood (and are thus labelled henceforth) as discourses, constellations of attitudes or values; in Dryzek's words: 'shared way [s] of apprehending the world' (Dryzek, 2005, p. 9).

| D ISCOUR S E S ON P OVERT Y WITHIN CONS ERVATION
Three discourses were identified, which we name as follows: Discourse 1: needs-based pro-poor with ecocentrism; Discourse 2: rights-based pro-poor; Discourse 3: ecocentrism with rights commitments. Figure 2 presents the numerical characterizations of discourses, showing z-scores 5 and normalized scores (corresponding with positions in Figure 1) for each statement. These numerical characterizations of the discourses are based upon an 'ideal-type' Q sort, which represents the mean ordering of statements for respondent Q sorts associated with this discourse. We encourage readers to directly consult Figure 2 to interpret differences between the discourses, and readers can cross reference Figure 2 while reading the following results section which closely references statements.
This information is presented in more conventional Q methodology format in Table S2. In what follows, we describe and interpret the discourses, referring to Q statement numbers and their normalized score (position in Figure 1 grid) in parentheses (Q-#: normalized score), and marking distinguishing statements (ranked in a significantly different way in other discourses; Watts & Stenner, 2012), with an asterisk. In addition, we illustrate the discourses with qualitative data (italicized) derived from the Q sorting process (marked with a superscript Q ), and interviews (marked with a superscript I ).

| Discourse 1: Needs-based pro-poor with ecocentrism
Discourse 1 expresses joint ecocentric and pro-poor anthropocentric imperatives. In straddling what are often thought of as opposite poles of anthropocentrism and ecocentrism, the discourse is relatively nuanced. It is also distinguished by emphasizing human needs and interests rather than human rights and safeguards. The emphasis on needs, interests and livelihoods finds resonance with ideas around the 'environmentalism of the poor' (Guha & Martínez-Alier, 1997). It is F I G U R E 2 Statements with corresponding z-scores and normalized scores for each factor. z-scores are displayed for each factor, with Factor 1 in blue diamond, Factor 2 in green square and Factor 3 in red circle. Where statement is distinguishing for the factor (at p < 0.05), the symbol is filled. The right-hand column indicates where a statement is a consensus statement, otherwise it indicates whether it is distinguishing at p < 0.05, and for which factor. Note that statement 23 is not a consensus statement, nor statistically significant in distinguishing between factors. Normalized scores are given, corresponding to the positions in the Figure 1 Q methodology grid associated with 16 respondents, including one US-based representative and five of seven respondents from each of China and Nepal, and five of six from Uganda; it therefore has very strong representation from these country case studies. The gender balance broadly represents the sample. It is associated with a range of organizations with species-oriented and broader conservation interests.
Discourse 1  The discourse hence recognizes the role of poor people in protection (Q-16, above), but distinctively compared to other discourses, in some circumstances also regards poverty as a threat to biodiversity. This discourse more strongly promotes human well-being as a conservation goal than other discourses (Q-26*: −3), and considers a moral imperative in conservation organizations seeking to alleviate poverty (Q-27: −2). The discourse weakly prioritizes humans above non-human nature (Q-21: −1; Q-32: −1), suggesting the following caveated response to Q-21 is typical: Non-human nature is essential to the health of the planet Earth and to human survival. There are times when human desires must not be met in order to save the planet [I-12, international dataset Q ].
There is a related indication that the risk of species extinction might elevate the rights of non-human nature (Q-25*: +1). Ecocentric dimensions are also expressed in the strong salience of intrinsic values of biodiversity (Q-15: +4) and the human virtue of caring for nature (Q-2: +3). In Q methodology, ideas are salient when they feature towards the grid extremes (see Figure 1).
Relatively high salience is given to the SDGs in promoting propoor environmental governance (Q-10*: +3), and greater social equality is prioritized as the rationale for conservation organizations to promote poverty alleviation (Q-7*: +3). The needs and rights of future human generations are salient as something conservation organizations should champion (Q-19*: +3).

| Discourse 2: Rights-based pro-poor
Discourse 2 is pro-poor and strongly emphasizes the 'do no harm' principle, human rights and social safeguards. It is more closely associated with Discourse 1 than 3. It is associated with 10 respondents, five international, and including one from each of Uganda and Nepal, and three from Bolivia. The gender balance roughly represents the sample. Most Discourse 2 respondents are drawn from organizations with a focus broader than species, and four of five of the international respondents are drawn from organizational signatories to the CIHR.
Discourse 2 promotes anthropocentric conservation imperatives (Q-26*: −1) and is distinctively characterized by giving priority to do no harm (Q-4*; +4) standards in conservation and international human rights norms (Q-18: +4; Q-6*; +3). This is exemplified in the qualitative data: Reactions to Q-23 (−1) and Q-25* (0) (non-salient, but with a lower rank position in this than any other discourse) also signify the importance of human rights within this discourse.
In Discourse 2 ranks Q-20 (−1) about nature being used to alleviate poverty now, rather than preserved for future generations, more highly than others. This discourse puts overall less emphasis than other discourses on ecocentric conservation and intrinsic value (Q-15*: +1), and prioritizes human rights, but appears agnostic about the rights of non-human nature (Q-25*: 0): When asked to consider situations where conservation and poverty alleviation are in direct opposition, associated respondents tend to prioritize humans, although Q-32 is not highly salient.

| Discourse 3: Ecocentrism with rights commitments
Discourse 3 is ecocentric and articulates some pro-poor commitments in terms of social safeguards. It shares with Discourse 2 the importance of human rights and safeguards, and a critique of wealth as a driver of conservation threats. Some aspects of ecocentrism are shared with Discourse 1. It is associated with three respondents from the international sample and two from the Bolivian country case study.
It is associated with four male respondents, and one female. Among international organization respondents associating with this discourse, two of three represent more species-oriented organizations. The idea of poverty being the main threat to tropical biodiversity is non-salient (Q-22*: −1). Broadly, in terms of the anthropocentrism and ecocentrism dimensions, all discourses share some element of ecocentrism, with Discourse 3 being the strongest followed by Discourse 1 and Discourse 2, respectively. Discourse 3 is more willing to prioritize non-human nature in the case of extinction (Q-25*; distinguishes all), whereas Discourses 1 and 2 lean more towards anthropocentrism and Q-25 accordingly has low salience (+1*; 0*, respectively).

| Summary of convergence and divergence between the discourses
All have some commitment to anthropocentrism, although there is some contention as to whether human well-being should be an ultimate goal (Q-26*) of conservation. However, different discourses bring to bear different rationales, with Discourse 1 prioritizing the interests (Q-28, 29 and 30), needs and sufficientarian basis (Q-8 [marginally] highest for Discourse 1) (sufficientarianism being the idea that everyone has a right to a decent livelihood; see Lehmann et al. (2018) or Gosseries (2011)

F I G U R E 3 Comparison of discourses on five key dimensions of difference.
Discourses are compared on a simple ordinal scale, and accordingly should only be interpreted in relative positions to one another (for instance, D3 is more ecocentric than D1). Relative positions are described in the text in Section 3.4, which references the corresponding Q statement rankings for selected dimensions

| COMPARING THE D ISCOUR S E S ALONG S IDE P OS ITI ON S IN THE LITER ATURE
All of the discourses share some commitments to not solely an anthropocentric, but rather a more demanding set of pro-poor concerns. Thus, Adams et al.'s (2004)  The second and essentially Brundtland viewpoint of 'poverty is a critical constraint on conservation' (Adams et al., , p. 1147) was a matter of contention in these data, with an interesting geographical inflection. The extent to which poverty was conceptualized as a driver of conservation threats (Q-22) is a statement that distinguishes all of the discourses, and particularly Discourse 1 (+1) from Discourse 2 (−4). While it is not a strength of Q methodology to link discourses with demographics, it remains worthy of note that Discourse 1 is very dominated by representatives from China, Nepal and Uganda. It is also notable that respondents from the global north and Bolivia were more circumspect about poverty as a driver of conservation threats, marking a strong departure from Brundtland. It is also instructive to contrast these perspectives about poverty as a driver with those relating to wealth as a driver of threats. Discourse 2 is arguably the most critical of threat drivers with a basis in wealth, and critical commentary about wealth and overconsumption as drivers was very much present in the qualitative data, particularly among Discourse 2 and 3 respondents. These discourses were overwhelmingly dominated by respondents from the global north and Bolivia, whereas the discourse that is dominated by respondents from China, Nepal and Uganda was more likely to associate some drivers with poverty.
The preliminary indicative findings we present here could motivate further research to investigate these questions with a research design more appropriate for understanding links between perspectives and demographics. However, there appear to be two elements to try and explain in relation to this preliminary finding. First, we need to explain a possible move away from the conceptualization of poverty as a driver among respondents from the global north, in contrast to the logic of the Brundtland commission, which predominated at least during the 1990s (Duraiappah, 1998). This perhaps reflects the more recent realities of globalized consumption, along with emerging research demonstrating the importance of contemporary globalized drivers of biodiversity threats, originating in the global north (Barlow et al., 2018;Moran & Kanemoto, 2017). It may also reflect the traction of ideas of environmental justice, and the environmentalism of the poor in challenging the previously dominant narratives. Second, we must explain why respondents from the global north tended to be more critical of wealth and overconsumption as drivers, compared to Asian or African counterparts in this research. This may be as a result of respondents feeling more able to criticize the impacts of their own societies, or perhaps more aware of the growing evidence demonstrating the shifting origin of conservation threats.
Within the conservation community, it is contentious whether conservation organizations should contest the global economic order, which frames globalized patterns of wealth and consumption (Büscher et al., 2017;Corson, 2010). Adams (2013)  It was also noted that some organizations do, increasingly, extend the remit of traditional conservation foci by adopting stances around broader issues of consumption and climate change, for instance [e.g.
We also used the research to examine whether respondents applied an instrumental logic to poverty and conservation. It is important to understand how significant this logic is today, given its influence in the Brundtland report and throughout the 1990s (Duraiappah, 1998). With reference to the typology of Adams et al. (2004), this is an extension of the second viewpoint; if 'poverty is a critical constraint on conservation', an instrumental logic supports the conservation movement taking an interest in poverty because poverty reduction will lead to conservation gains.
While this logic was certainly evident at times in qualitative data, for instance: we care about people's wellbeing because we see it as a purposeful way to get to a conservation outcome.
So we see improving people's wellbeing as a way to provide incentives for conservation practice [I-10, in- and a further quote: I think we have to be totally up front and say that as an organisation our mission is very clear: to save species from extinction. So… we work with human communities as a way of saving species from extinction… You can't do one without the other [I-07, international However, despite some instrumental perspectives being present in the qualitative data, statement Q-14, which explicitly tests an instrumental logic, was not salient for any discourse (Figure 2). Indeed, the placement of Q-14 (and furthermore the pragmatism embodied in Q-11, which is also non-salient), points to a normative-rather than instrumental-logic that respondents applied to the Q methodology instrument. This normative logic is also very evident in the following qualitative data (in direct response to Q-14): I also disagree with this, because there is the moral obligation, no? [I-35, Bolivian dataset Q ]; There is a moral obligation to consider human outcomes, at a minimum in order to do no harm, but ideally to actively promote human welfare [I-03, international dataset Q ].
In contrast, therefore, to the instrumental perspective, the predominant view, particularly within Discourses 2 and 3, was an ethically justified idea that 'conservation activities must, at the very least, Indeed, and going further into rights issues, many respondents associated with Discourse 2 highlighted in interviews that the realization of rights can in fact promote conservation.
While ideas about rights were strongly associated with Discourse 2, and also prominent in Discourse 3, one important finding of this research was that they had little resonance in Discourse 1. This is evident in a group of statements which most clearly allow Discourses 1 and 2 to be differentiated (Q-4; Q-6; Q-18; Figure 2).
While Discourse 1 takes a welfare-oriented approach to poverty, Discourse 2 (and 3 to an extent) prioritizes more of a rights-based approach. That this emphasis on rights is not, however, universally prioritized, is an important indicative finding of this research. It is however important to question the significance of this, given that safeguards and rights are often designed specifically to protect welfare and needs. Yet, the communication of these aspects was consistently distinct in different discourses, and at the very least, the mismatch might lead to a different basis for claims to natural resources, different policy imperatives and challenges of communication between those associated with different discourses. It is also worth highlighting that beyond the corollary emphases of rights and needs, there remains substantive disagreement on Q-4 and the acceptability of harm to humans from conservation.
Another implication of this differentiation in discourses is geographical. The difference between Discourse 1 (largely respondents from China, Nepal and Uganda) and, particularly Discourse 2, but also Discourse 3 (both largely respondents in the global north) on the 'do no harm' position and the emphasis placed upon rights and safeguards in harm mitigation is significant. It may even signify implications for the potential for shared perspectives between conservation organizations working in partnership across the global north and south. One respondent discussed the challenges of harmonizing the implementation of social policies across a large con-

| CON CLUS ION
We investigate contemporary discourses about poverty and conservation within the conservation sector and offer the first empirical analysis of perspectives on this nexus. Using a novel combination of Q methodological and more conventional qualitative analysis, Motivations for conservation are often characterized as dichotomously either anthropocentric or ecocentric (e.g. Kareiva & Marvier, 2012), and this has been a problematic facet of the 'new conservation' debates (Holmes et al., 2017). However, our analysis shows the complex hybridity of perspectives within our sample, with all three discourses containing some elements of both. This supports recent evidence that contemporary conservationists adhere to both imperatives (Sandbrook et al., 2019). However, more effectively than a Likert survey, Q methodology does reveal respondents' ultimate priorities, for instance in response to statements 21 and 32, which convey the essential ecocentrism of Discourse 3 ( Figure 2). Discourses 1 and 2 lean most towards anthropocentrism, but they are characterized by different orientations of conservation to poverty. Hence, below the surface of the 'anthropocentric turn' we traced above, there appear to be two essentially distinct framings of poverty considerations that conservation practitioners adhere to. The first prioritizes welfare, needs and sufficientarianism and this perspective is more strongly associated with respondents in China, Nepal and Uganda. Our research design does not enable us to generalize these findings, but further research could investigate whether these perspectives are associated with the continents of Asia and Africa. The second framing of poverty drawn upon by Discourses 2 and 3 focuses much more on do no harm principles and social safeguards, and this appears to follow a western, or internationalized human rights-oriented discourse. It is striking to note, therefore, that while Discourse 3 is essentially ecocentric, it was not accompanied by a disregard for poverty concerns, but instead an adherence to this internationalized rights discourse. This suggests that agendas such as the Conservation Initiative on Human Rights may have traction even among organizations dominated by ecocentric perspectives.
Using Q methodology, underpinned by more nuanced qualitative research, we have elicited and elaborated upon areas of consensus and divergence within conservation practitioner perspectives. We expect that these results and analyses will serve to update, parse and clarify perspectives on poverty within the conservation movement, illuminating consensual aspects and revealing where critical differences remain. Although the three discourses lean in different directions on anthropocentrism and ecocentrism, there is significant common ground on many principles of both. There is consensus that the poor should not shoulder the costs of conservation, but two distinct framings of poverty emerge, bringing different emphases and implying distinct policy imperatives. Finally, practitioners afford different weights to poverty and wealth drivers of environmental threats, and debates are clearly ongoing among our respondents and elsewhere (Adams, 2013), as to what this changing picture of drivers demands of the conservation, and broader environmental, movement.

ACK N OWLED G EM ENTS
We thank the research participants who gave very generously of their time, and offered fascinating insights. We thank the two reviewers for detailed and rigorous reviews which enabled us to improve the paper and we thank C. Guiver for generous assistance in the preparation of Figure 2. The early part of this work was shaped in discussion with Prof. Thomas Sikor, to whom we are deeply grateful. This research was supported by the Ecosystem Council.

CO N FLI C T O F I NTE R E S T
We do not have any conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The Q methodology data underpinning this study are publicly available from the UK Data Service at http://resha re.ukdat aserv ice.ac.uk/, with https://doi.org/10.5255/UKDA-SN-854313 (Fisher, 2020).

O RCI D
Janet A. Fisher https://orcid.org/0000-0001-6549-2358 E N D N OTE S 1 We reviewed website pages from organizations within the sample for the dual purpose of developing Q statements and preparing interviews. We selected the most appropriate pages (for instance, regarding work with local communities, or explicitly about approaches to poverty), and undertook qualitative thematic coding.
while emphasizing that these can be contradictory, and holistic judgement is required in order that extracted factors are meaningful in Q methodology terms. The Eigenvalues (or Kaiser-Guttman) criterion would have led us to extract six factors. However, this would not represent much reduction of the correlation matrix, and Brown (1980) argues that this criterion is relatively meaningless in Q studies. A further criterion is that meaningful factors should have at least two associated respondents. This would eliminate the four and five factor solutions. Humphrey's rule (see Watts & Stenner, 2012) would suggest a two-factor solution was appropriate. Overall, we drew on our experience of Q methodology and also considered the accompanying qualitative dataset, which indicated that the third factor contributed meaning to the study, and aided understanding in terms of contrast with the other two factors. Including it meant that ecocentrism was expressed in the Q analysis, and because this was very evident in the qualitative data, we considered it merited expression through Q. In summary then, the extraction of the third factor was supported by the eigenvalue criterion, the criterion that a meaningful factor must have at least two associated respondents (Factor 3 has 5, and explains 13% of study variance). Its inclusion was also supported by related qualitative data, although it did not meet the Humphrey's rule criterion.
4 Factors were rotated using a varimax procedure within PQMethod, followed with 'by-hand' rotation (this combination is described by Watts & Stenner (2012, p. 126) as useful and effective for exploiting the 'complementary strengths' of both processes). We rotated F1 and F2 anticlockwise by 2 degrees to incorporate two further respondents in the factor solution (moving from eight to six respondents not associated with any factor). This greater incorporation of respondents in factors is noted as desirable by Stricklin and Almeida (2000) and Watts and Stenner (2012). Respondents were flagged for factors using manual flagging (as recommended by Watts and Stenner (2012), when exceeding the threshold of 0.46 and a clear 10 decimal points higher than their association with another factor; see Watts & Stenner, 2012). 5 z-score numeric values are given in Table S2.
6 It should however also be noted that this quote was later coupled with an assertion about it being unacceptable for the poor to bear the burden of conserving public goods.