Who knows, who cares? Untangling ecological knowledge and nature connection among Amazonian colonist farmers

1. Conservationists often assume that connection with and caring about nature's well- being is strongly linked


| INTRODUC TI ON
Severe declines in regional biodiversity are accompanied by the simultaneous loss of knowledge about natural ecosystems within traditional and non-traditional human societies (Aswani et al., 2018;Miller, 2005).
Loss of ecological knowledge raises concerns that it may undermine both our ability to manage ecosystems and our interest and willingness to protect them. Decades of research in conservation psychology and environmental education have demonstrated that the relationship between knowledge, attitudes and pro-environmental behaviours is complex and nonlinear (Kollmuss & Agyeman, 2002;Rickinson, 2001;Varela-Losada et al., 2015;Wals et al., 2014). However, many conservationists-most of them trained in natural sciences (Bennett et al., 2017)-still believe that a sense of connection and caring about nature is based on knowledge and understanding of the natural world, particularly the knowledge of different species (Table 1).
farmers are capable of forming strong connections with nature, even if they rarely possess detailed knowledge of local forest biodiversity. Considering the complex and apparently context-dependent relationship between knowing and caring about nature, it is unwise to assume that changing one would automatically affect the other.

K E Y W O R D S
Amazon, birds, conservation psychology, ecological knowledge, farmers, nature connection TA B L E 1 Examples of the belief that caring for nature is related to ecological knowledge expressed in environmental NGO materials and academic papers in conservation science

Quote Context Source
NGO examples 'In the end, we will conserve only what we love, we will love only what we understand, and we will understand only what we are taught' Quote by Baba Dioum (1968) used as inspiration in an educational poster campaign for schools, WWF UK One Planet Schools 'Learn' Programme WWF UK Webpage (2008) 'We approach our elephant work in the same spirit as Baba Dioum. Our endeavour to protect elephants is inspired by love and deep respect for them as a species and as individuals, which, in turn, is based on our understanding of them gained through long-term study' ElephantVoices, quoting Baba Dioum (1968) at the top of the Education section of their webpage ElephantVoices Webpage (2018) Academic examples 'People care about what they know' Study demonstrating that UK children had greater identification knowledge of synthetic Pokémon 'species' than common British wildlife Balmford et al. (2002) ' [T]he levels of ecological knowledge studied here (names of living components of ecosystems and the functions and uses of each component) provide an indication of a community's connectivity and willingness to care for the local environment, since naming things with which we are familiar is human instinct and we are unlikely to care about that which we do not know' (p. 1007) Cross-cultural study demonstrating an inverse relationship between levels of ecological knowledge and wealth measured at the level of nation and community Pilgrim et al. (2008) ' [The] loss of familiarity and knowledge [of nature] is cause for profound concern as it may lead to reduced appreciation of the natural world, reduced motivation to protect species, [and] less willingness to support nature (…)' Study exploring socio-demographic factors influencing plant identification knowledge among adults in UK Robinson et al. (2016) 'People who care, may make choices to conserve; but people who don't know [nature], don't even care. What is the extinction of a condor or an albatross to a child who has never known a wren? ' (p. 207) Opinion piece emphasising the need for fostering deep connections to nature Pyle (2003) 'Natural history, the scientific study of plants and animals in their natural environments, is the cornerstone of ecological literacy.
It not only instructs in the knowledge of place but instils an emotional enthusiasm and empathy toward natural phenomena' (p. 118) Study highlighting deficient levels of natural history knowledge among university students in Mississippi, including those with course work in ecology Hammond and Herron (2012) Crossing the sub-disciplinary divides of applied ecology and environmental psychology, the literature on the relationship between ecological knowledge of species and nature connection includes substantial gaps. Studies explicitly examining this link are few, the findings are mixed, and the evidence is limited to richer Global North countries. Despite long-standing recognition that conservation hinges upon the knowledge and cooperation of local people (Berkes, 2004), there is little information on the levels of ecological knowledge and nature connection among people living in proximity to many important biodiversity areas in the Global South. One significant specific gap surrounds levels of ecological knowledge and nature connection among non-indigenous colonist farmers who migrated to farm-forest frontiers, where they are important local conservation stakeholders (Campos & Nepstad, 2006). It is also unclear whether ecological knowledge and nature connection have the same or different individual-and landscape-scale determinants, and thus the extent to which their development may be interdependent.
We address these gaps by investigating relationships between ecological knowledge of wild bird species and psychological connection with nature-and their drivers-among colonist farmers living at a major deforestation frontier in the Brazilian Amazon. We test the correlation between ecological knowledge and nature connection, and we identify and compare their geographical, socio-demographic and experiential correlates among local farmers.

| Theoretical underpinnings of nature connection and ecological knowledge
Psychological nature connection is a multidimensional construct defined as the extent to which a person self-identifies with nature (cognitive connection, also known as connectedness; Schultz, 2002) and the extent to which a person feels emotionally attached to nature (called emotional or affective connection; Kals et al., 1999;Perkins, 2010;Schultz, 2002). According to the Inclusion of Nature in Self theory, the stronger someone's nature connection is, the greater is their propensity to empathise and feel concerned for nature's well-being (Schultz, 2002;Tam, 2013b). Thus, nature connection explicitly triggers the biospheric motivation for nature protection, concerned with nature for nature's sake (Schultz, 2001). Numerous measures of nature connection, emphasising different aspects of this multidimensional construct, have been shown to be strongly statistically convergent and underpinned by the same latent general variable (Tam, 2013a). This means that results from studies using different measures related to nature connection can be meaningfully compared under a single framework.
Familiarity with local species and the ability to name them is a component of 'ecological knowledge', defined as the cumulative body of knowledge, practice and beliefs concerned with sitespecific interactions between living beings (including humans) and between organisms and their environment (Berkes, 1999;Olsson & Folke, 2001). Ecological knowledge that is accumulated, evolved, and culturally transmitted over generations is termed 'traditional ecological knowledge' (TEK). Groups such as colonist farmers may lack this cultural and historical continuity of interactions with their (new) environments; nonetheless, they can build 'local ecological knowledge' (LEK), based on some mixture of practical knowledge, media and peer learning (Olsson & Folke, 2001 Berkes, 1999). The propensity to name and categorise organisms is evident cross-culturally; humans appear to have an innate system primed for the recognition of life forms and their ordering into taxonomies (Medin & Atran, 2004). Thus, the ability to name species, though not exhaustive of ecological knowledge, can be regarded as its fundamental form-the first layer of familiarity with one's ecosystem.

| Evidence for and against a link between ecological knowledge and nature connection
Several studies indicate that ecological knowledge relates positively to nature connection and other related concepts, such as 'environmental sensitivity' or positive attitudes towards birds. For example, Hammond and Herron (2012) found that university students in Mississippi with higher self-reported 'environmental sensitivity' ('having empathy for or relating to other living things or nature in general', p. 120) were more knowledgeable about identification and natural history of local fauna and flora than their peers. Cox and Gaston (2015) showed that the extent to which British people liked different bird species and 'felt connected to nature when watching birds in their gardens' was positively related to their bird identification skills. Finally, White et al. (2018) found that bird identification knowledge and positive attitudes towards birds among British schoolchildren were positively correlated, although changes in attitudes and changes in knowledge following a 6-week bird feeding and monitoring programme were unrelated.
However, Lumber et al. (2017) found very different results in a UK-based psychological study using a validated nature connection measure in online surveys and a quasi-experimental intervention based around a nature walk in a university campus park (Lumber et al., 2017). Their results suggested that knowledge-based activities were ineffective at increasing nature connection, which was instead enhanced by activities based on finding contact, emotion,

| Drivers of nature connection and ecological knowledge
Ecological knowledge and nature connection are rarely examined together, and it is unclear whether they are influenced by the same factors. Nonetheless, both are theoretically and empirically recognised as rooted in people's unique personal experiences of nature (Clayton et al., 2017;Miller, 2005;Soga et al., 2016;Turvey et al., 2010).
Drawing on the largely separate literatures on ecological knowledge and nature connection, we can identify groups of environmental, social and geographical factors that may shape them simultaneously.
These groups include what we here refer to as 'nature-contact',  Kals et al., 1999;Soga et al., 2016). There is also evidence of a positive feedback loop because people who visited natural spaces in childhood and those with higher nature connection are more likely to continue visiting natural areas in adulthood (Lin et al., 2014;Rosa et al., 2018). Formal education can also be classified as a nature-contact factor, emerging from the ecological knowledge literature. Education has been frequently (but not always) linked to losses of traditional ecological knowledge among indigenous populations (Aswani et al., 2018, p. 2) because it often reduces the time that children spend in natural areas, either alone or with knowledgeable elders (Demps et al., 2015). The effect of education on nature connection remains less explored.
Nature-reliance factors affect the extent of people's direct material reliance on nature to satisfy basic needs such as food, shelter and medicine. These factors relate to broad-scale socio-environmental transformations including modernisation (spread of technology, urbanisation, modern health services), market integration and growing wealth. These transformations are the principal candidate explanations for declining ecological knowledge in traditional populations (Aswani et al., 2018;Pilgrim et al., 2008), which act by changing people's lifestyles and by lowering people's direct reliance on natural ecosystems and thus the need for intimate knowledge of the local environment. Pilgrim et al. (2008) also suggested that increases in societal wealth are likely to similarly negatively affect whether people care about nature, but this hypothesis remains largely unexplored.
Nature-access factors relate to the accessibility of nature experiences. Declines in biodiversity, natural habitats and restrictions to accessing remaining natural areas, for example, former common lands, have all been linked to losses in ecological knowledge, presumably through decreased opportunities to interact with nature (Barreau et al., 2016;Kai et al., 2014;Miller, 2005). There is also evidence that the 'extinction of experience' resulting from physical separation from nature, most extreme in many cities, may lead to disconnection from nature among urban dwellers, especially children (Miller, 2005).
For example, in Stockholm, children attending preschools located close to areas offering diverse nature experiences were more empathetic towards the natural world than children from preschools with lower access to such areas (Giusti et al., 2014).
Beyond factors related to contact, reliance and access to nature, differences in ecological knowledge and nature connection have been associated with socio-demographic factors such as gender, age and culture (e.g. Aswani et al., 2018;Luck et al., 2011), although the directions of these associations are context-dependent.

| Nature connection and ecological knowledge at deforestation frontiers
It is important to understand the relationship between nature connection and ecological knowledge in the Global South, particularly at tropical forest-farm frontiers where biodiversity loss is often most intense (Barlow et al., 2018). Many such frontiers are characterised by remoteness and relatively weak state presence and are inhabited by colonist farmers without traditional ties to the land, and whose individual decisions collectively shape the landscape and hence conservation outcomes (Fearnside, 2008).

| Study aim and research questions
Here we assess the relationship between ecological knowledge and nature connection in the Transamazon Highway region, a mature Amazonian forest-farm frontier. We address this main aim by asking two research questions. First, is there a positive relationship between ecological knowledge and nature connection among Transamazon colonist farmers? We achieve this by comparing two validated, independent measures of nature connection with two purposely developed measures of ecological knowledge based on the ability to identify local forest and non-forest bird species.
Second, do ecological knowledge and nature connection share common drivers? This is answered by examining the roles of access to, contact with and reliance on nature. In addition, we assess how knowledge of birds and nature connection varies according to social and demographic characteristics including geographical origin, gender and age. Because we make use of observational data and cannot infer causality, we talk about predictors and associations rather than drivers and effects when discussing our results. We sample various farms across the region to capture diverse experiences of nature, which we expect to relate to high levels of variation in remaining forest cover, urban accessibility and household reliance on forest foods.

| Study site
Our study area is located around the Transamazon Highway, in the south-eastern Brazilian Amazon in Pará state (see Figure 1)

| Study design and questionnaire application
Sampling was stratified to capture variation in local forest cover and distances from the sub-regional urban centre Altamira (

| Ecological knowledge measures
Our ecological knowledge measures were based on the ability to recognise local bird species. We chose birds as they are ubiquitous, are commonly used in research on ecological knowledge of species, many are relatively easy to identify for non-specialists and most have little utilitarian value locally (few species are regularly hunted).
Therefore, we consider bird identification skills a good proxy for general ecological knowledge of species identification. Additionally, based on our previous observations from the Amazon basin, we suspected that the colonist farmers will be more familiar with species commonly observed near human settlements and in heavily anthropogenically modified habitats than forests. To maximise the chances of identifying knowledgeable 'experts' but also differentiating between people at the lower end of knowledge levels about bird species, the species in our sample were chosen to (a) represent both forest and non-forest habitat, (b) be relatively common locally (Lees et al., 2013a, A. C. Lees unpublished abundance data for the neighbouring region of Santarém, Pará) and (c) be comparably easy to identify in the field without binoculars. The selected species are also not commonly hunted or persecuted, although some species are prized as cage birds.
Participants were asked to name species from two bespoke plates with photographs of 19 birds: 13 non-forest species that occupy agricultural areas and six forest-associated species (Table S2).
To account for the possibility that some birds may be easier to recognise by sound than by sight, we then played calls of a sub-sample of eight species (five non-forest and three forest) with easily recognisable songs and calls and asked the participants to name the species.
Lastly, to account for the possibility that respondents may recognise species but not necessarily know their common names, we asked the participants to match the recorded calls to the images. Thus, we obtained three complementary-but not fully independent-measures of bird recognition. Some names proffered by interviewees did not exactly match the target species but could be matched to closely related species or higher-level classifications, roughly corresponding to genus and family level. To account for this, names were scored for correctness at species and approximately genus and family level. See Supporting Information Section S3 for further details on ecological knowledge measures construction and scoring.

| Socio-demographic and nature-experiencerelated factors
The 'natureza') and in opposition to pasture, which some even referred to as 'deserts'. Importantly, the term 'forest' appeared inclusive of both primary and regenerating forests, as well as agroforestry plantations.  Factors were below 2, as assessed using the imcdiag function in the mctest package ver 1.2 (Imdadullah et al., 2016). Relatedness between nature connection and ecological knowledge measures was assessed using Spearman correlations; significance was tested using t tests with Holm correction for multiple tests (alpha = 0.5).

| Analysis
Generalised linear models with beta-binomial error distribution and logit link were used to test the associations between sociodemographic and nature-related factors as predictors and LCNR, INS, forest bird ecological knowledge and non-forest bird ecological knowledge as outcome variables. The models were run on complete cases data (n = 227) using the package gamlss ver. 5.1 (Stasinopoulos & Rigby, 2007). Continuous predictors were standardised by centring and dividing by two standard deviations and the binary predictors (gender, origin) were centred (rescale function in package arm). A predictor was considered significantly associated with an outcome variable if the 95% confidence interval of the estimated beta-coefficient did not contain 0.
Beta-binomial models are appropriate for modelling data in the form of a discrete number of successes in a fixed number of trials with unknown probabilities. In our models, each point that was The OR for an x-unit change in a predictor is calculated by raising the beta-coefficient exponent to the power of x. For the standardised predictors in our models, ORs for a one-unit change on the original scales were calculated by diving the beta-coefficients by two standard deviations before taking the exponents.
Additionally, having found much inter-species variation in recognition rates, we ran a post-hoc mixed-effects logistic regression for the probability of recognising a bird species from an image (n = 4,408 observations), exploring the effects of species body mass (EltonTraits, Wilman et al., 2014), habitat (forest/non-forest), our subjective assessment of ease of visual detection (high, medium and low) and relative abundance (high/low) based on a survey in Santarem (Lees et al., 2013b, unpubl. results) which was separated for forest and non-forest species. We used the species (n = 19) and respondent id (n = 232) as random effects.

| Bird identification knowledge
On average, respondents recognised 46%-61% of non-forest bird species and 12%-15% of forest species at the genus level ( Figure 2; and S2). Therefore, for subsequent analysis, we used the sum of recognition scores for images, sounds and sound-image matching, with image and sound recognition scored at the genus level, separately for forest and non-forest species. This produced two ecological knowledge measures (one for each habitat) that were less biased towards any single recognition method, although more strongly weighted towards those species which were included in all three methods.
The post-hoc analysis of interspecific visual recognition rates revealed strong effects of habitat, body mass and relative abundance ( Figure 4, see Table S3 for model statistics). All else held equal, the odds of correct identification were 9.53 times higher for the locally common versus rare species, 8.43 times higher for non-forest versus forest species and increased 1.17 times for a 100 g increase in body mass.
In summary, respondents generally held more knowledge of local non-forest bird species and poor knowledge of forest-associated species, with larger and more common species recognised more often than others.

| Associations with socio-demographic and nature-experience-related factors
There was no similarity between the correlates of ecological knowledge of birds and nature connection ( Figure 6; see Tables S4 and S5.

| D ISCUSS I ON
Contrary to a common conservationist discourse (Table 1), our survey of Amazonian colonist farmers across diverse landscapes at an Amazonian deforestation frontier found no evidence that people with greater ecological knowledge of species feel more strongly connected to nature. Moreover, knowledge and connection did not share any predictors in common, indicating that they are shaped by different social processes. These findings were consistent across the two measures of nature connection we employed (cognitive Inclusion of Nature in Self and emotional Love and Care for Nature-Rural scales) and across our two measures of ecological knowledge-one for non-forest species and other for forest-associated bird species. We also found that Transamazonian farmers tend to have high levels of nature connection, moderate ecological knowledge of non-forest bird species and very low ecological knowledge of forest bird species; they also recognise large and abundant birds more often than others.
Farm-forest landscapes are critical arenas in which tropical habitats and biodiversity are declining (Barlow et al., 2016), yet they have been largely overlooked by conservation psychology research (Mastrangelo et al., 2014). In these Amazonian settings, where traditional forest peoples have been largely pushed out to reserves, with much territory lost to agriculture expansion, colonists have come to be recognised as key 'conservationist actors' (Campos & Nepstad, 2006). We show that the colonist farmers can develop a relatively strong connection with F I G U R E 6 Generalised linear model results for nature connection and ecological knowledge of bird species, showing the standardised beta coefficients of predictor variables with their 95% confidence intervals nature, even without traditional ties to the land or detailed knowledge of local biodiversity. However, their capacity for protecting the local environment may be constrained by poor knowledge of forest taxa (Peterson et al., 2008), particularly in the case of small, rare and inconspicuous species which form the bulk of the forest-associated fauna.
More generally, our findings, which contrast with previous studies from Europe and the United States, imply that the link between ecological knowledge of species and nature connection is context-dependent and highlight the need to better understand the specific mechanisms that shape each one of them.

| When is knowing nature related to caring about nature?
We show that the positive relationship between ecological knowledge and nature connection, found in the United States and the United Kingdom, is not universal. We found no correlation between knowledge and connection; most people felt strongly connected to nature, but their knowledge of bird species varied widely ( Figure 5). Notably, even some of the few people who were expert at identifying birds had below-average nature connection.
Although our ability to detect relationships may have been impaired by a possible ceiling effect (i.e. saturation) in the affective nature connection measure, the tendency for high scores on nature connection measures is not uncommon (e.g. Cox & Gaston, 2015).
Moreover, the independently measured cognitive connection was also unrelated to either measures of bird identification knowledge, suggesting that the lack of significant relationship between ecological knowledge and nature connection is not merely a statistical artefact.
Our findings are congruent with recent psychological research exploring what types of nature experiences promote a holistic, committed relationship with nature among adults and children (Giusti et al., 2018;Lumber et al., 2017;Richardson & McEwan, 2018 (Cox & Gaston, 2015;Hammond & Herron, 2012;White et al., 2018). Where does this discrepancy come from?
The association between ecological knowledge and nature connection is inevitably influenced by the choice of measures. Social psychology shows that behaviour-specific attitudes correlate with pro-environmental behaviour much more closely than general proenvironmental attitudes (St. John et al., 2010). Consequently, we might expect that object-specific measures of ecological knowledge (e.g. bird identification skills) will correlate more closely with measures indicating connection with specific elements of nature (e.g. attitudes towards birds) than with measures of general nature connection, such as those we used.
To assess nature connection across people with different levels of ecological knowledge, the choice of taxa used to estimate this knowledge also matters. We show that species with attributes such as large body size, high abundance and ease of observation are easier to identify than others, highlighting the importance of including species with diverse traits to distinguishing experts from less knowledgeable people. Our selection of birds was purposively skewed towards species that we expected to be easily recognised. It proved well suited for our study population, with good differentiation between people across the scale of ecological knowledge. In populations with higher average knowledge, the inclusion of 'harder' taxa, such as invertebrates and small mammals (Medin & Atran, 2004), could improve the ability to discriminate between individuals with high levels of ecological knowledge. Conversely, including 'easy' taxa like large mammals could extend the lower end of the scale where necessary.
The contrast between our research in rural Amazonia and the results from US-and UK-based studies suggest that the relationship between ecological knowledge and nature connection varies with the socio-cultural and ecological context. We cannot rule out that ecological knowledge and nature connection may in some conditions stimulate one another. However, our results also suggest that knowledge and connection are shaped by separate processes, shown by the apparent lack of common predictors and opposing associations with age: positive for connection and negative for knowledge. This conclusion is supported by the contrasting effects of modernisation and wealth in other studies. For example, in a cross-cultural study across India, Indonesia and UK, these factors associated with lower ecological knowledge and its concentration among a few specialists (Pilgrim et al., 2008), but in a longitudinal US study, these same drivers promoted an intergenerational shift towards seeing wildlife more as human-like and deserving of care (Manfredo et al., 2020). Hence, we caution against assuming that ecological knowledge and nature connection are co-dependent. Instead, we recommend focussing on understanding the mechanisms that drive them.

| Who is connected with nature?
Even though many Amazonian agriculturalists equate cultivation and conversion of natural land with progress and success (Hoelle, 2018), Furthermore, the apparent lack of a relationship between nature connection and forest cover suggests that the importance of physical access to nature is secondary to social factors that shape whether and how people interact with it.
We found nature connection increases during adulthood, supporting previous studies (Hughes et al., 2019). A possible explanation might be a gradual increase in the cognitive and emotional bond between a person and place (Baldwin et al., 2017).
Older colonist farmers have typically lived in Amazonia and on their properties for longer than younger counterparts, providing greater opportunities for attaching to nature on their properties (Bogdon, 2016). We found no evidence of a nature-connection gender gap in Amazonia, contrasting with previous work in the United States and China (Schultz & Tabanico, 2007;Tam, 2013b We did not find support for Pilgrim et al.'s (2008) assertion that nature connection is positively related to nature reliance.
Neither the relative dependency on forest foods (low on average, Figure S3) nor distance to the nearest market town, presumed to indicate reliance on subsistence foods, related to nature connection.
Unexpectedly, however, we found that people living nearby to the sub-regional city Altamira reported somewhat higher emotional nature connection than those living farther away. Perhaps this reflects a spatial gradient in the activity and influence of environmentalist NGOs, government regulatory bodies, academic researchers and agricultural extension services. These institutions generally champion the view of small and medium landholders as custodians of the local environment (Schwartzman et al., 2010;de Toledo et al., 2017) and most of them operate out of Altamira. Farmers may respond to these interactions over time with a change in their underlying nature connection, and/or changes in the way they talk about the environment with outsiders (i.e. a potential reporting bias for psychological surveys).
Overall, it appears that nearly 50 years of forest loss and socioenvironmental transformation in eastern Amazonia have not resulted in any clear 'extinction of experience' and a resultant loss of nature connection, counter to the tendency in the Global North (Miller, 2005).

| Who knows about nature?
Transamazonian farmers appear to have low-to-moderate ecological knowledge, placing them somewhere on a spectrum between highly knowledgeable traditional peoples and the industrial societies where ecological knowledge is generally low but skewed by knowledgeable specialists (Medin & Atran, 2004;Pilgrim et al., 2008). Amazonian colonists arrived with little prior knowledge of the Amazonian environment (Moran, 1977). Some of the bird species living in the Amazon are the same or similar to species found in other parts of Brazil, so migrant farmers could have had pre-existing knowledge of them. However, in the case of Amazonian endemics, this knowledge is presumably built mainly through the farmers' own, sometimes inter-generational, experiences. Whereas shapes and behaviours can be learned from observation alone, names are harder to learn, given they are normally acquired from other knowledgeable people and sometimes media.
Indeed, we found that for some birds, respondents could correctly match their images and sounds, but did not always know their names (Table S1).
The social basis of ecological knowledge acquisition was reinforced by our finding that knowledge of forest and non-forest birds was related only to socio-demographic factors, but not to nature-access or nature-contact measures. However, our contact measures focused only on the frequency of visits to natural habitats but did not account for time spent in agricultural areas, which could relate specifically to knowledge of non-forest birds, so this gap remains to be explored. The two factors that related to ecological knowledge were age and gender. Younger farmers were more knowledgeable about non-forest bird species than older respondents, resonating with a study of migrant farmers in Sumatra (Nyhus et al., 2003). We also found that men were far better than women at recognising bird species. This gendered gap in ecological knowledge is context-specific and has been found also in the United States, United Kingdom and the Narok region in Kenya (Huxham et al., 2006;Kassilly, 2006;Kellert et al., 1987), but not in the Netherlands (Hooykaas et al., 2019) or the Nkuru region in Kenya (Kassilly, 2006). The gendered differences in ecological knowledge are likely explained by the division of labour and resulting differences in exposure to nature, since men generally spend more time outside engaged in farming, agroforestry and occasionally hunting, whereas women traditionally tend to engage in home-bound activities (Oestreicher et al., 2014).
We did not find a positive linkage between reliance on nature and ecological knowledge, in contrast with previous work (Aswani et al., 2018;Pilgrim et al., 2008). Though most of our respondents reported some use of forest products, their livelihoods were rarely dependent on them ( Figure S3). There were also no linkages of ecological knowledge with proximity to the nearest urban area or the regional town. Potentially, the associations with nature-reliance measures may have been stronger had we measured ecological knowledge of species of utilitarian value, such as that of medicinal plants or game species.
Some birds were much easier for the colonists to identify than others. Larger and more abundant species were better known than smaller and rare species, resonating with previous studies from Germany and China (Kai et al., 2014;Randler et al., 2007). Another important factor influencing recognition was habitat. Consistent with the results of Nyhus et al. (2003) on mammals, people were much better at recognising non-forest species than the forest ones. This is unsurprising because most of the non-forest species in our sample are widespread throughout Brazil, so migrant farmers could learn to identify them even before coming to Amazonia and could pass this knowledge to their children. Second, in the case of species novel to them, the colonists likely learn to recognise non-forest species first. Farmers tend to spend more time in pastures and plantations than in old-growth forests, and visually detecting elusive forest species is often impossible without binoculars. Nonetheless, poor recognition of forest species implies that farmers may underestimate forest biodiversity and the damage resulting from forest conversion and disturbance, potentially constraining their capacity for conservation (Peterson et al., 2008;Pollock et al., 2015).

| CON CLUS IONS
We show that colonist farmers living along a farm-forest frontier in the Brazilian Amazon tend to be strongly connected to nature, but their knowledge of local forest biodiversity remains limited.
Contrasting with studies from the Global North, we found no evidence that ecological knowledge and nature connection are related to each other or shaped by the same socio-demographic and natureexperience-related factors. Therefore, we caution against the common assumption of co-dependence between ecological knowledge and nature connection. Where they are considered important for conservation, we recommend focussing on understanding the context-specific mechanisms that drive each one of them. Although willingness and knowledge are both important for nature protection, we must remember that farmers' decisions are also shaped by wider systemic forces, including markets, social norms and the law. To achieve effective and just conservation, the relevant policies, norms and regulations must be harmonised to incentivise pro-conservation practices both socially and economically so that people may feel capable of protecting the environment and maintaining other opportunities to lead lives they have reason to value.

CO N FLI C T S O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Anonymous data supporting the conclusions of this manuscript have been archived in the Dryad Digital Repository, available at https://doi.org/10.5061/dryad.08kpr r520 (Mikołajczak et al., 2021).