Coexistence in rivers of hunger: Human–otter relations and “Shamans of the Waters” in an Amazonian socioecological system

1 Introduction

Human-wildlife relationships are not static. Throughout history, shifting social meanings of wildlife have shaped the roles animals occupy within different societies (Frank and Glikman, 2019). Historically, many anthropocentric conservation perspectives separated humans from nature and wildlife (Oliver and Johnston, 2000), either physically—through cities, infrastructure, or protected areas—or symbolically, by assigning wildlife to spaces conceived as distinct from human life, such as biological reserves. In Amazonia, however, many conservation models—including Indigenous Lands and Sustainable Development Reserves—are characterized by the permanent presence of resident human populations. In these contexts, boundaries between human and wildlife spaces are more fluid and are actively constructed through local social values, cosmologies, and everyday practices (Knight, 2013; Philo and Wilbert, 2000; Creager and Jordan, 2002). These relational boundaries define how particular societies negotiate coexistence with non-human animals, rather than enforcing strict spatial segregation.

How human societies perceive wildlife strongly influences the interactions established between humans and non-human animals, the space fauna occupies within social landscapes, and whether these relationships are characterized by coexistence, neutrality, or conflict (Knight, 2013; Philo and Wilbert, 2000; Creager and Jordan, 2002). Social theory further suggests that shared in-group values shape attitudes, preferences, and behavioral intentions toward wildlife and its management (Sponarski et al., 2015). Consequently, human–wildlife interactions must be understood as context-specific, multidimensional, and embedded within broader socioecological systems that integrate environmental resources with the social identities and worldviews of human groups (Frank and Glikman, 2019; Lute and Gore, 2019).

The experience of the Baniwa Indigenous people in Northwestern Amazonia and their long-standing relationship with otters (Pteronura brasiliensis and Lontra longicaudis) offers an instructive model for understanding complex human–predator dynamics in subsistence-based societies. The Baniwa refer to otters as “Shamans of the Water”, and consider them guardians of fish populations in lakes, which provide key protein resources in this low-productivity environment (Pimenta, 2016). This cosmological understanding, rooted in stories and practices going back centuries, continues to inform contemporary Baniwa relationships with aquatic environments and non-human animals.

This relational system, however, was profoundly disrupted during the mid-20th century. From the 1920s through the 1970s, Baniwa commercial hunters became involved in trade networks supplying global demand for wild animal furs for the clothing and fashion industry. The “Shamans of the Water” were transformed into valuable commodities, their skins hunted and sold to regional, non-Indigenous middlemen (Antunes et al., 2016; Pimenta et al., 2018a). This period, regionally remembered as the “Age of Skins” (in Portuguese, Era das Peles), led to severe population declines of Neotropical otters and the near-local extinction of Giant Otters within Baniwa territory (Pimenta et al., 2018a). Especially with the arrival of Protestant missions to Baniwa territory in the 1940s, Baniwa shamans also became increasingly rare, coinciding with the decline of the lake-protecting “Shamans of the Water”.

Following the enactment of Brazil’s Fauna Protection Law in 1967 and the country’s ratification of CITES, Baniwa people today comment on the coincidence between the local disappearance of otters in the 1970s (Pimenta et al., 2018a) and a noted decline in fish populations beginning during the same period. In 1975, otter populations began to recover across Amazonia (Uscamaita and Bodmer, 2010). However, in the upper Rio Negro, clear local signs of recovery only emerged in the 1990s, after the demarcation of Indigenous Lands and the expulsion of outside traders from the region (Pimenta et al., 2018b). Following legal protection and the decline of commercial hunting, otter populations gradually recovered in the upper Rio Negro. Contemporary Baniwa narratives frequently associate this recolonization process with broader transformations in aquatic environments and fishing dynamics (Goulding et al., 1988; Moran, 1991).

Human populations often coexist with wildlife, sharing space and resources without necessarily framing these interactions as conflictual. Competition becomes manageable, guaranteeing the coexistence between human beings and various animal species (Frank, 2016; Nyhus, 2016; Hill et al., 2017). However, in systems where humans and wildlife exhibit strong overlaps in ecological niches, competition for shared resources can approach or exceed the environment’s carrying capacity, increasing the likelihood of conflict and undermining the feasibility of coexistence (López-Bao et al., 2015).

The Neotropical otter and the Giant Otter are semi-aquatic mustelid carnivores. Both species are sympatric across much of their range (Duplaix, 1980; Carter and Rosas, 1997), reflecting ecological conditions that allow co-inhabitation (Groenendjik et al., 2005; Silva et al., 2013). The establishment and persistence of the otter populations are linked to the availability of habitat and fish resources, as well as unpolluted and unimpeded water bodies that favor locomotion and prey capture (Staib, 2005; Lima et al., 2012; Pimenta et al., 2018b). These same environments are widely used by Indigenous and riverine populations, for whom fishing is often the main source of animal protein (Cerdeira et al., 1997; Begossi et al., 2018). This apparent overlap in habitat and resource use has led otters—especially Giant Otters—to be widely perceived as competitors by fishermen throughout the Amazon basin (Gómez and Jorgenson, 1999; Recharte et al., 2008; Michalski et al., 2011; Rosas-Ribeiro et al., 2012; Lasmar and Marmontel, 2013; Lima et al., 2014). Such perceptions frequently translate into conflictual relationships and direct persecution, which, together with habitat loss, remain among the principal threats to otter conservation in the region (Duplaix et al., 2015).

Although conflict-oriented interpretations of human–otter interactions are commonly reported across many Amazonian contexts, such patterns should not be uncritically generalized. From a strictly ecological perspective, conflict might be expected to intensify in the upper Rio Negro— a “river of hunger” system characterized by low primary productivity and limited fish availability (Goulding et al., 1988). However, interspecific dynamics are not determined by ecological constraints alone, but are also shaped by cultural values, social norms, and historically grounded practices. Consequently, understanding human–otter coexistence in Amazonian Indigenous territories requires a context-specific socioecological analysis that integrates both ecological and cultural dimensions.

Building on the need for context-specific socioecological analyses, here we examine spatial and trophic relationships among Giant Otters, Neotropical Otters, and Baniwa fishers on the upper Rio Negro using a multidisciplinary approach that integrates ecological data and Indigenous knowledge. We assess patterns of habitat use and diet composition to evaluate the extent of niche overlap, while explicitly focusing on the cultural drivers—values, cosmology, attitudes, and practices—that shape human–otter interactions. By jointly considering ecological constraints and culturally grounded drivers of behavior, this study advances understanding of coexistence in an oligotrophic Amazonian system and highlights how Indigenous socioecologies can simultaneously support food security and the recovery of historically exploited otter populations.

2 Materials and methods

2.1 Study area and cultural context

The upper Rio Negro is located in the northwest Amazon within the Brazilian municipality of São Gabriel da Cachoeira (Figure 1), along the border with Colombia and Venezuela. The region is home to a huge diversity of indigenous peoples, including around 20 ethnic groups speaking languages belonging to five distinct cultural-linguistic families: Tukano, Arawakan, Maku, Yanomami, and Tupi-Guarani (Cabalzar and Ricardo, 2006).

The Baniwa people, and their closely related neighbors in Colombia, the Koripako, belong to the Arawakan language family and have inhabited the Içana River Basin for many centuries (Wright, 2005). The Içana River Basin originates in Colombia, but most of its 696 km of length lies within Brazilian territory. In Brazil, the Içana River runs through the Indigenous Land of the Upper Rio Negro, receiving water from several tributaries (including the Ayari, Cuyari, Piraiawara, and Cubate rivers), until it joins the Rio Negro (Cabalzar and Ricardo, 2006).

Throughout its course, the Içana River is characterized as a black water river system surrounded by low-productivity white-sand savannas. Local variations in water level, soil type, and human habitation history create a diverse mosaic of landscapes and vegetation types (Abraão et al., 2010), with floristic composition significantly shaped by past and ongoing Baniwa management activities (Franco-Moraes et al., 2019). The middle Içana, at the confluence with the Rio Ayari, is dominated by seasonally flooded igapó forests on sandy soils. Though extremely poor in nutrients and mostly inappropriate for agriculture due to annual flooding, this region contains numerous blackwater lakes that local people value for their comparatively abundant fish populations (Shepard et al., 2004). Such lakes also provide good habitat for otters (Pimenta et al., 2018b), which made the region a very attractive place for otter hunters and traders to supply the international fur market during the 20th century (Pimenta et al., 2018a).

This region comprises the traditional territory of the Dzawinai or “jaguar people,” one of several patrilineal Baniwa clans. This “lakes region”, as the Baniwa call it, represents our study area. It covers around 65 km of river extension and is home to 10 Baniwa communities. We visited nine of these communities for this study: Jandú Cachoeira (1), Tucumã (2), Bela Vista (3), Urumutum Lago (4), São José do Ayari (5), Arapasso (6), Tucunaré Lago (7), Tarumã (8), and Santa Marta (9).

2.2 Data collection

2.2.1 Perceptions, attitudes, and behavior toward otters

We conducted semi-structured interviews with Baniwa residents to document perceptions, attitudes, historical memory, and behavioral observations regarding otters and fishing activities in the middle Içana River. Interviews addressed four main themes: (1) fishing practices and preferred fishing grounds; (2) interactions between otters and fishing activities; (3) historical changes in otter populations, including commercial hunting during the fur trade period and subsequent recolonization; and (4) perceptions and narratives associated with otter recovery and coexistence.

Interviews followed a flexible script combining open-ended and guided questions, allowing participants to elaborate on personal experiences, local ecological knowledge, and intergenerational observations. Participants included adult residents (≥18 years old) from multiple Baniwa communities distributed throughout the study area. Basic demographic information, including age, gender, occupation, and length of residence in the community, was also recorded.

2.2.2 Baniwa preferences

In addition to the semi-structured interviews, we conducted 17 free-listing interviews across seven communities, divided into five zones shared among family groups, to identify preferred fishing locations and fish species consumed during the dry season. The free-listing method consists of asking respondents to provide lists of items associated with a given topic, allowing the identification of shared perceptions and culturally salient concepts (Borgatti and DSH, 1999). The number of respondents by area was included in the subsequent analyses to standardize the sampling effort under the number of quotes about the preferred fishing locations, given that the area covered by each family group is delimited on an established territorial zoning by the indigenous organizations themselves.

Each interviewee could list five preferred water bodies for fishing and 10 preferred fish for food. Both questions were restricted to the low-water period, which coincides with the ecological sampling period used for subsequent niche overlap analysis (see 2.2.2). The fish species were presented in the native language, and the Portuguese correspondence was made by the indigenous and non-indigenous specialist authors (J.Z. and V.M.) with the support of the Baniwa fish catalog “Kopenai Nako - about fishes” (Pamáali, 2006). Following food item identification, we determined each fish taxon’s family and size. This level of classification allowed direct comparability with other studies that have used data extracted from otter feces to facilitate subsequent data analysis.

All data collection activities were carried out in close consultation with, and active participation by, a team of Baniwa researchers and knowledge keepers led by co-author Valêncio Walipere Macedo.

2.2.3 Neotropical and giant otter niches

To estimate otter habitat use, we surveyed 150 km of watercourses, including 19 lakes (97.3 km) and 16 streams (52.6 km), over 22 consecutive days at the beginning of the low-water period (September to November), when otters are restricted to these habitats. Surveys were conducted between 06:00 and 18:00 h from a boat travelling at a maximum speed of 10 km/h (Yoccoz et al., 2001; Pollock et al., 2002; Groenendjik et al., 2005), with the entire shoreline of each water body inspected for direct records (sightings of individuals or groups) and indirect signs (e.g., footprints, latrines, burrows, and resting sites). All survey routes were tracked using a Garmin eTrex 10 GPS, and each otter record was georeferenced. Surveys were carried out with Baniwa researchers and knowledge holders, whose familiarity with the area improved the detection of otter signs and sightings. For each water body, records were compiled separately for Giant and Neotropical Otters and used as abundance indices in subsequent analyses.

To estimate the Neotropical and Giant Otter diet, we collected fecal samples collected during the watercourse surveys. The ichthyological content of the samples was identified based on the protocol established by (Rosas et al., 1999).

2.3 Data analysis

2.3.1 Qualitative analysis

Interview notes were systematized and analyzed qualitatively through thematic content analysis, focusing on recurring themes, shared perceptions, and convergent narratives related to coexistence, resource use, behavioral adaptation, and environmental change. Rather than quantifying responses statistically, the analysis aimed to identify common socioecological patterns and culturally grounded interpretations.

2.3.2 Salience index

From the lists provided by the Baniwa fishers during the application of the free listing interview, we calculated the Salience Index of the responses. The SI combines the frequency with the average position of each term mentioned, to verify the cultural consensus established among the interviewees for the explored theme (Sutrop, 2001). SI index values can range from 0 (no citations) to 1 (cited by all participants). This allowed us to rank preferred Baniwa fishing locations and preferred species for consumption during the dry season in the middle Içana River.

2.3.3 Habitat use overlap

To test whether Baniwa preferred fishing locations which overlapped with habitats used by otters, we modelled the salience index of each sampled water body as a function of otter abundance and habitat type using generalized additive models for location, scale, and shape (GAMLSS). The response variable was the Baniwa salience index for each water body, while predictor variables were Giant Otter abundance, Neotropical Otter abundance, habitat type (lake or stream), and the number of interviews conducted in each area to control for differences in sampling effort among zones.

Before modelling, we assessed collinearity between predictors using pairwise correlation tests and considered variables with |r| ≥ 0.60 to be collinear. No problematic collinearity was detected. We then compared alternative distributions for the response variable and selected the Gumbel distribution (GU) based on the generalized Akaike information criterion (GAIC) and residual diagnostics. After choosing the response distribution, we fitted a candidate set of models, including null, additive linear, and additive non-linear formulations. Non-linear relationships were tested using cubic spline terms (cs) and penalized B-splines (pb).

Models were compared using GAIC, and the model with the lowest information criterion was retained as the final model (Burnham and Anderson, 2004). To visualize the effects of predictors, we plotted partial effects and 95% confidence intervals from the final model.

2.3.4 Diet overlap

To assess the extent to which fish preferred by Baniwa fishers overlapped with those consumed by Giant and Neotropical otters, we compared the composition of fish among the three groups. For all analyses, fish taxa were standardized at the family level, allowing direct comparison between interview-based and fecal-sample data. We used the proportional representation based on the number of species reported/detected within each fish family in each sample as the input matrix. Before analysis, values were square root transformed to reduce the influence of highly abundant taxa while retaining information on relative abundance. Dissimilarities among samples were then calculated using the Bray–Curtis index. Based on this dissimilarity matrix, we performed a Principal Coordinates Analysis (PCoA) using classical multidimensional scaling and plotted the first two axes for visualization. To aid interpretation of group separation in ordination space, convex hull polygons were used to represent group boundaries, as they provide a non-parametric representation of sample dispersion and avoid assumptions of multivariate normality. The proportion of variation explained by each axis was calculated from the eigenvalues. Differences in diet composition among groups (Giant otter, Neotropical otter, and Baniwa) were assessed using Permutational Multivariate Analysis of Variance (PERMANOVA) with 999 permutations. Because PERMANOVA is sensitive to differences in within-group dispersion, we tested for homogeneity of multivariate dispersion, followed by permutation tests (999 permutations). Post-hoc comparisons of dispersion among groups were performed using Tukey’s HSD.

All statistical analyses were conducted in R (R Core Team, 2023) within the RStudio environment (Posit Team, 2023). GAMLSS was conducted using the gamlss package. Multivariate analyses, including Bray–Curtis dissimilarity, PERMANOVA, and tests of multivariate dispersion, were performed using the vegan package. Graphical visualizations were produced using ggplot2 and ggrepel packages.

3 Results

3.1 Baniwa knowledge and human-otter interactions

We conducted semi-structured interviews with 19 Baniwa residents across seven communities in the middle Içana River. Interviewees ranged from 24 to 88 years old, and all participants were male fishers, reflecting the traditional gendered division of labor in Baniwa society, where fishing and hunting activities are predominantly performed by men. Most respondents reported living in the same community for several decades, reflecting extensive familiarity with local lakes, fishing practices, and long-term changes in otter populations. The main themes emerging from the interviews are summarized in Table 1.

Table 1. Main themes identified in semi-structured interviews with Baniwa fishers regarding coexistence with otters in the middle Içana River, northwestern Amazonia.

Most interviewees (84%) reported frequent overlap between otters and fishing activities during the dry season, especially in blackwater lakes that constitute preferred fishing grounds. However, only a minority of participants (32%) explicitly described otters as competitors for fish resources or as negatively affecting fishing activities by dispersing fish schools. Rather than framing these interactions primarily as conflictual, most respondents described coexistence strategies based on avoidance and mutual distancing.

When encountering Giant or Neotropical otters during fishing activities, 63% of participants reported behavioral adjustments, including temporarily leaving the area, moving to another part of the lake, waiting for the otters to leave, reducing noise, or avoiding direct approaches. Several interviewees explained that disturbing otters during fishing activities is considered inappropriate or unnecessary, particularly in lakes where otters are frequently observed. As one fisherman from the community of Tucumã Rupitá explained, “they belong to that place.”

A recurring theme among interviewees was the description of Giant Otters as elusive animals that avoid human presence by retreating when hearing paddles or boat engines, a behavior mentioned by 65% of respondents. Several participants contrasted this behavior with historical accounts describing otters as less fearful in the past. Approximately 37% of interviewees explicitly associated this behavioral change with the commercial hunting period linked to the fur trade, which some respondents interpreted as otters having “learned” to avoid humans following decades of persecution.

Participants frequently associated Giant Otters with the protection and regulation of fish populations, reinforcing the Baniwa notion of otters as “Shamans of the Water.” Associations between otters and fish recovery were mentioned by 68% of interviewees, many of whom described otters as guardians, owners, or “kings” of fish and lake environments. Respondents commonly recalled the decline or disappearance of otters during the fur trade period and linked this process to broader ecological degradation, including reductions in fish abundance. Conversely, the recolonization of otters after the end of commercial hunting was widely perceived as part of a broader process of environmental recovery and restoration of traditional territorial dynamics. Detailed descriptions of the commercial hunting period and Baniwa origin narratives associated with otters are presented in Pimenta et al. (2018a).

Importantly, when discussing declines in fish abundance, interviewees more frequently associated them with increasing human population density, intensified fishing pressure, and the introduction of predatory fishing techniques such as timbó poisoning and extensive gillnet use, rather than with otter predation.

3.2 Habitat use

The Baniwa cited 30 water bodies used as preferred fishing grounds in the middle Içana River during the low-water season. Lakes (19) are more often used than streams (11); the 10 water bodies with the highest salience among Baniwa fishermen are also lakes. Three of the water bodies used by the Baniwa could not be visited for ecological sampling, and therefore were excluded from subsequent analyses - Mawiriana (SI = 0.18, lake), Kadzali (SI = 0.08, stream), and Dowiria (SI = 0.33, stream).

Another eight water bodies were sampled that were not considered by the Baniwa as fishing grounds. These areas were treated as having a salience index (SI) of 0.0 and included in the analysis for a total of 35 sampled water bodies.

The Baniwa’s choice of fishing locations was associated with water body type and the abundance of otter species (Table 2). Baniwa people showed a clear preference for lakes as fishing grounds (Figure 2A). Giant Otters abundance was negatively associated with the salience index (SI) of habitat use by fishers (p= 0.0179) (Figure 2B), whereas Neotropical Otter abundance showed a positive association with SI (p= 0.0464), indicating strong overlap in habitat use between fishers and small otters (Figure 2C).

Table 2. Selected model from the generalized additive model for location, scale, and shape (GAMLSS) (AIC = -7.33, family of distribution = Gumbel, link function = identity) assessing the relationship between the salience index of habitat use by Baniwa fishermen with abundance of giant otter and neotropical otter, and type of water body in the middle Içana River, Amazonas, Brazil.

Significance values are shown *p <0.05, *** p<0.001, and “.” indicates a weak effect.

3.3 Fish consumption by otters and the Baniwa people

The Baniwa cited 20 preferred fish for consumption during the dry season. The salience index shows a preference for consuming mid-sized to large fishes from the Cichlidae, Anostomidae, Pimelodidae, and Serrasalmidae, respectively. By contrast, Giant Otters prefer mid-sized to small Cichlidae, large and small Erythrinidae, and small Doradidae. Neotropical Otters prefer small Doradidae, Cichlidae, Erythrinidae, and other non-identified species (including some macroinvertebrates, such as crabs, beetles, and mollusks) among the most consumed items (Table 3).

Table 3. Diet overlap.

The five most representative diet items of each evaluated group are highlighted.

Salience index of Baniwa’s fishery diet during the dry season in the middle Içana River and the frequency of food items identified in the feces of otters collected during the same period.

When assessing the frequency of food items present in otter feces, the only fish group appearing in first place for both the Baniwa and the Giant Otter as the most consumed prey was medium-sized cichlids. Surprisingly, dietary overlap was greater between Giant and Neotropical otters than between either species and humans, focusing especially on small-sized species from Doradidae and Cichlidae (Figure 3). The first two axes of the Principal Coordinates Analysis explained 36.1% of the total variation in diet composition (PCoA1 = 23.9%, PCoA2 = 12.2%). In ordination space, Baniwa samples formed a relatively tight and distinct cluster, clearly separated from otter samples along the first axis. In contrast, samples from Giant and Neotropical otters were more widely dispersed and showed partial overlap, particularly along the second axis.

The PERMANOVA confirmed marked differences in diet composition among the three groups (F₂,₃₃ = 13.33, p = 0.001, R² = 0.447). This result demonstrates that nearly 45% of the multivariate variation in prey composition is attributable to differences among Baniwa, Giant otter, and Neotropical otter diets, indicating a strong structuring of diet according to group.

The contrast between Giant otters and Neotropical otters was statistically significant (Pairwise PERMANOVA: F₁,₁₇ = 4.04, R² = 0.192, p = 0.005), although the effect size was relatively modest, with only 19% of the variation explained. In contrast, comparisons involving Baniwa revealed much stronger differentiation. Diet composition differed strongly between Giant otters and Baniwa (Pairwise PERMANOVA: F₁,₃₀ = 21.01, R² = 0.412, p = 0.001), with over 41% of the variation explained by group identity. An even greater separation was observed between Neotropical otters and Baniwa (Pairwise PERMANOVA: F₁,₁₉ = 16.71, R² = 0.468, p = 0.001), where nearly 47% of the variation in diet composition was attributable to group differences.

Analysis of multivariate dispersion revealed significant heterogeneity in within-group variability (ANOVA: F₂,₃₃ = 6.94, p = 0.003; permutation test: p = 0.007), indicating that the spread of samples in multivariate space differed among groups. Post-hoc Tukey comparisons showed that Baniwa exhibited significantly lower dispersion than both Giant otters (difference = −0.166, 95% CI [−0.307, −0.025], p = 0.018) and Neotropical otters (difference = −0.280, 95% CI [−0.502, −0.059], p = 0.011), whereas no significant difference in dispersion was detected between the two otter species (difference = 0.114, 95% CI [−0.110, 0.338], p = 0.435).

4 Discussion

Niche segregation minimizes competition for resources when these resources exist in limited abundance in the shared environment, if the species explore distinct microhabitats or adopt different behavioral patterns associated with foraging strategies. In the study at Jaú National Park (lower Rio Negro) (Silva et al., 2013), the competition for fish resources between Giant and Neotropical otters was mitigated by preference for different prey sizes. In addition, in the southern Pantanal, the overlap of resources used by the same two otter species was mitigated by the abundance of food resources (Muanis, 2008).

By contrast, our study suggests that the low productivity of the aquatic environments in the upper Rio Negro’s tributaries places greater dietary restrictions on predators. Giant Otters maintained a predominantly piscivorous diet (Duplaix, 1980; Rosas et al., 1999; Kruuk, 2006; Rosas-Ribeiro et al., 2012), while Neotropical Otters followed a more opportunistic pattern with a predominance of fish consumption but frequent inclusion of small vertebrates and macroinvertebrates (Muanis, 2008; Kruuk, 2006; Larivière, 1999; Rheingantz and Trinca, 2015; Rheingantz et al., 2012). Nonetheless, in contrast to other studies conducted in areas of otter species overlap, the prey consumed by the two otter species show some similarities, including small Cichlidae and Erythrinidae species and, especially, small Doradidae. All are commonly found in shallow margins of lakes and streams and in places with sunken logs and stands of aquatic herbaceous plants.

Unusual partial dietary overlap between the two otter species observed in the middle Içana appears strongly influenced by the Giant Otters’ adoption of a more generalist diet, with a broader size range of prey items, including those more typically consumed by Neotropical Otters (e.g. small Doradidae and Characidae). This modification in the Giant Otter diet may reflect the species’ high energy demands [individuals of the species can consume 10% of their own body weight in fish per day (Carter et al., 1999)] in a low-productivity environment, potentially fostering competition with Neotropical Otters. However, this pattern should be interpreted with caution, as the low sample size for Neotropical Otters limited the range of identifiable food items and may have underestimated the breadth of their diet.

Nonetheless, this apparent competition for food resources is mitigated by the spatial segregation of foraging areas, since Neotropical Otters mostly use lakes while Giant Otters seek their prey in shallow streams. Small streams (up to 10m width and 3m depth) and forest creeks represent good habitats for Giant Otters (Pimenta et al., 2018b) because of the ease of encountering and capturing fish for food (Staib, 2005; Lima et al., 2012), whereas lakes provide structurally complex environments that may favor the more opportunistic foraging behavior of Neotropical Otters. Therefore, differential use of habitats likely promotes coexistence of these two mustelid species by reducing competitive interactions.

However, coexistence in the middle Içana River cannot be explained solely by ecological segregation. Interview data further reinforces that coexistence is also maintained through socially embedded behavioral practices. Although Baniwa fishers widely recognized spatial overlap with otters in preferred fishing lakes, most respondents did not frame otters as direct competitors. Instead, interviewees emphasized mutual avoidance, reduced disturbance, and behavioral adjustments by both humans and otters as mechanisms that facilitate coexistence within shared aquatic environments.

Animal behavior also contributes to the construction of coexistence landscapes (Lamb et al., 2020). Carnivores are particularly known for their ability to reduce their home range in anthropized regions (Tucker et al., 2018); by increasing nocturnal activity to avoid encounters with humans (Carter et al., 2012; Gaynor et al., 2018); and to adjust their feeding habits, either by developing new foraging specializations to mitigate competition (Tinker et al., 2009; Ribas et al., 2012), or by becoming more generalist in the face of new opportunities (Battin, 2004; Lamb et al., 2017). They can also transmit part of these learned social behaviors to their offspring. In a study carried out around urbanized areas in Canada, grizzly bears born from “non-problematic” mothers (vis-à-vis their interactions with humans) were more likely to avoid conflicts with humans than those born from “problematic” mothers, indicating that there is matrilineal transmission of preferences in bears’ attitudes towards humans (Morehouse et al., 2016).

According to comments by Baniwa knowledge keepers, Giant Otters shy away as soon as they hear an engine or even a paddle hitting the water. For them, this elusive behavior -- unusual for this territorial species (Duplaix, 1980; Carter and Rosas, 1997) -- reflects Giant Otters intention to avoid interactions with humans, an behavioral adaptation which in turn may have resulted from the threats posed by commercial hunting in the past. In this sense, Baniwa observations reinforce the results of quantitative habitat use analyses by suggesting that Giant Otters may have behaviorally adapted to historical hunting pressure in ways that continue to reduce direct interactions with humans today.

This interpretation is reinforced by interview data, in which 65% of respondents described Giant Otters as elusive animals that avoid boats and human presence, while 37% explicitly associated this behavior with historical hunting pressure during the fur trade period.

Importantly, when discussing fish declines, interviewees more frequently associated declines in fish abundance with increasing human population density, intensified fishing pressure, and predatory fishing techniques such as timbó poisoning and gillnet use, rather than with otter predation. These finding challenges generalized assumptions that human–otter interactions in Amazonia are primarily structured through direct competition for fish resources, highlighting the importance of historical, social, and behavioral dimensions in shaping coexistence within shared aquatic landscapes instead.

Beyond these empirical patterns, the Baniwa case also speaks to broader theoretical debates on human–wildlife coexistence. Recent debates on human–wildlife coexistence have called for approaches that move beyond species-centered and conflict-oriented frameworks, emphasizing instead the relational, historical, political, and more-than-human dimensions through which coexistence is produced (Van Bommel and Boonman-Berson, 2022; Krauss et al., 2025). Our findings contribute to this discussion by showing that Baniwa–otter coexistence emerges from the interaction of ecological differentiation, historical memory, behavioral adaptations by both humans and otters, relational norms, and changing governance conditions. Importantly, incorporating Baniwa knowledge reveals dimensions of coexistence that would remain largely invisible through conventional ecological approaches alone.

Within Baniwa cosmology, agency is distributed among humans, Giant and Neotropical otters, fish, and the Shamans of the Water, who is understood as a protector and regulator of aquatic resources. Otters are likewise perceived as active participants in maintaining fish populations and adapting their behavior in response to environmental and social conditions. Oral histories associated with the fur trade period and episodes of fish scarcity, together with local experiences of changing commercial regimes and hunting regulations, further demonstrate how historical and institutional transformations continue to shape contemporary human–otter relations. Recognizing these more-than-human relationships allows for a more nuanced understanding of how coexistence is produced, maintained, and negotiated.

In this sense, the Baniwa case demonstrates the value of Indigenous knowledge systems for advancing coexistence research by revealing ecological, cosmological, historical, and relational processes that underpin human–wildlife interactions and challenge conventional nature–culture divides in conservation science. More broadly, these findings suggest that a fuller understanding of coexistence emerges when Indigenous and scientific knowledge systems are engaged as equally rigorous and mutually informative ways of understanding socioecological dynamics.

5 Final considerations

Despite the severe environmental limitations of a region historically described as a “river of hunger” and often portrayed as a biologically impoverished and socially limiting blackwater system (Goulding et al., 1988), coexistence persists in the upper Rio Negro, challenging the assumption that human–wildlife relations in oligotrophic environments are necessarily constrained by scarcity and competition. The recovery of otter populations following their historical decline due to commercial hunting (Pimenta et al., 2018a) further suggests that coexistence is sustained by long-standing relationships among humans and non-human beings who share landscapes and resources over time.

Crucially, the historical breakdown of coexistence in the upper Rio Negro resulted from the imposition of an exogenous economic system, embodied by the international fur trade, which generated extractive pressure and disrupted Indigenous cosmologies, territorial practices, and relational boundaries between humans and non-human beings. The subsequent recovery of otter populations suggests that coexistence is not a fixed condition, but a dynamic process that can be eroded, negotiated, and restored through changing social, political, ecological, and cultural circumstances. More broadly, our findings reinforce previous studies, showing that low ecological productivity does not result in low biodiversity in the Rio Negro basin (Beltrão et al., 2019). Biological diversity and coexistence emerge from the complex and long-term interaction between environmental constraints and highly specialized adaptive practices of Indigenous and riverine peoples, who actively shape socioecological systems through territorial mobility, ecological knowledge, and culturally embedded forms of resource management, improving and shaping the composition of aquatic and terrestrial ecological systems.

Our study shows that conflict between humans and mustelid top predators in the upper Rio Negro is neither ecologically inevitable nor culturally intrinsic. These findings contribute to ongoing debates in conservation social science and human–wildlife coexistence research by showing that coexistence cannot be reduced to patterns of resource use, spatial overlap, or conflict avoidance alone. Likewise, it cannot be understood as the product of Indigenous cosmology or cultural values alone. The Baniwa case shows that coexistence requires attention to historical memory, behavioral adaptation, governance arrangements, and more-than-human relationships that shape how humans and wildlife share landscapes and resources. By situating human–wildlife interactions within a broader socioecological system, this study highlights the value of relational approaches that recognize the multiple human and non-human actors involved in the production and maintenance of coexistence.

Conducted with the active participation and coordination of Baniwa researchers in dialogue with their practices and concepts, this study demonstrates the power of collaboration and exchange between conventional science and Indigenous epistemologies and cosmologies to promote richer, socially and historically grounded, and ultimately more effective conservation science.

Future research that brings ecological, ethnographic, and participatory approaches (e.g. fishing diaries and harvested fish) into closer dialogue may further advance understanding of how coexistence is maintained, transformed, or challenged across Amazonian socioecological systems. Comparative studies in regions where otters are more frequently perceived as competitors may also help identify the ecological, historical, and social conditions that shape different pathways of coexistence.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Chiara Bragagnolo, Federal University of Alagoas, Brazil

Reviewed by: Geoffrey Garver, McGill University, Canada

Silvio Marchini, University of Oxford, United Kingdom