2.1 Literature Search

We conducted a literature review using the search string ‘(“Cerdocyon thous” OR “crab-eating fox”) AND “diet” AND “fruit” AND (“seed dispers*” OR “frugivor*”)’ in databases of Google Scholar, Science Direct, Scopus and Web of Science during the period of search from August 2022 to May 2023. To be included in the review, studies needed to (i) present any of the search string terms in their title, abstract or keywords, and (ii) convey information about C. thous diet and its potential for seed dispersal (e.g., fruit species consumed, foraging habits, habitats where seeds are dispersed).

Besides English, we included studies in Portuguese and Spanish, the primary languages of the countries where C. thous occurs, to ensure broad coverage of relevant sources and capture regionally significant studies that might otherwise be overlooked. Additionally, we did not restrict our search by publication year, allowing the inclusion of older or less widely known studies that could contribute valuable local knowledge. Duplicated studies were excluded. Dissertations and theses were also considered as long as they had not generated similar-content papers already in our search. A complete list of the 37 finalist studies included in our study is available in Table S1.

2.2 Frugivory

Based on the studies obtained in the literature review, we described the frugivorous diet of C. thous by retrieving all fruit species consumed by the canid. Plant species whose parts were consumed for reasons other than frugivory, for example, digestive aid, accidentally—mainly Poaceae species (Motta-Junior et al. 1994; Rocha et al. 2004, 2008)—were not considered as part of the fruit diet of C. thous. Unidentified fruits or generic plant groups (e.g., grasses, miscellaneous) were also not considered. All plant species names were verified for accuracy and current taxonomy using Plants of the World Online (POWO) and List of Species of the Flora of Brazil—REFLORA to ensure correct spelling and valid nomenclature.

We considered studies performed in native habitats (i.e., those mainly composed of natural vegetation), modified areas (i.e., anthropised habitats to some degree but still with some natural vegetation) and controlled environments (e.g., laboratory, captivity). We also recorded the seasons covered in the reviewed studies to assess the seasonality of fruit consumption by C. thous. These features provided thorough data across various timeframes and ecological contexts.

We used the Frequency of Occurrence (FO) to quantify the importance of fruits in C. thous diet (Klare et al. 2011). Typically, FO is expressed as a count of the number of samples (e.g., scats, stomachs) that contain the food item (e.g., fruits, insects, birds, small mammals) out of the total number of samples examined, indicating whether such an item is relatively common in the species' diet (Dietz 1984; Motta-Junior et al. 1994; Morin et al. 2019). FO is one of the most widely adopted techniques for determining carnivores' diet (Klare et al. 2011), enabling comparisons among studies, particularly relative to diet changes in time and space (Brzeziński and Marzec 2003). We obtained FO from 29 studies of the literature review.

FO only accounts for the presence of a certain food item among samples, disregarding its biomass (Porto and Rui 2019). It is preferable to use biomass calculations for better evaluating diet composition in carnivores (Klare et al. 2011). Nonetheless, just a few studies in the literature (n = 11) used methods besides FO while describing C. thous diet, let alone appraising biomass (n = 3/11) or volume (n = 4/11). Thus, due to the low sample size, we opted not to present analyses with the alternative metrics of Percentage of Biomass and Percentage of Volume found in the literature.

As an additional diet metric, we included data solely on the Percentage of Occurrence (PO), a derivative metric of FO. PO shows the proportion of samples in which a food item is present relative to the total number of samples, expressed as a percentage (Pedó et al. 2006; Bianchi et al. 2014). As samples often contain remains of multiple dietary items, FO expressed in percentages can sometimes exceed 100% (Newsome et al. 1983; Morin et al. 2019). While FO gives a raw count of how often an item is found in the samples, PO standardises this measure, facilitating the comparison across different studies or food items. To estimate the frequency of a given fruit species in C. thous' diet, we compared the number of its records in the literature review with the total number of studies analysed.

2.3 Seed Dispersal

For all consumed species, we compiled information on fruit and seed size (both measured as width) from the literature (M. A. Pizo unpubl. data). Fruits that were only identified at the genus level were assigned a size by averaging across all congeneric species available in the database. Whenever available, we compiled information regarding the seed dispersal performance of C. thous (e.g., foraging behaviours, deposition sites, dispersal distance, seed viability and effects on germination). From studies that evaluated seed germination (n = 8), we considered whether C. thous exhibited any effect on germinability (i.e., the ability of seeds to germinate) after seed ingestion and the effect (positive, negative or neutral) on germination percentages compared to a control test (seeds collected directly from fruit).

3.1 Literature Review

We retrieved 37 studies from the literature addressing aspects of the C. thous frugivory (Table S1). Studies varied widely in duration, ranging from short-term observations lasting 4 days to extensive, long-term studies spanning up to 11 years (132 months; Table S2). The average study length was 34.7 months (±37.6). Although studies covered different Neotropical biomes (Atlantic Forest, Cerrado, Pampa, Caatinga, Amazon, Pantanal, Chaco, Páramo; Table S2), they were predominantly conducted in Brazil's central-southern region (Figure S1), most of them from the Atlantic Forest (44.7%), Cerrado (21.1%) and Pampa (13.2%) biomes (Figure 1). The dominant study site type was native habitats (52.6%), followed by modified areas (42.1%) and controlled environments (e.g., captivity; 5.3%). Our analysis, therefore, encompasses a wide variety of environments with distinct levels of disturbance (e.g., native vegetation, crops, pastures, highways) and physiognomies (e.g., forests, savannas, marshes, grasslands).

To determine the C. thous diet, a large proportion of studies used faecal analysis (62.8% studies; sample sizes: 5–301 scats), while under a fifth sampled stomachs (16.3% studies; sample sizes: 2–46 roadkill foxes). Other studies adopted distinct methods, such as germination tests (16.3% of studies), seed removal experiments (2.3% of studies) and camera traps (2.3% of studies) (Table S2).

Most studies that addressed seasonality (n = 22 studies; Table S2) spanned both rainy and dry seasons, whereas just a small proportion focused exclusively on the dry season (n = 1 study; Table S2). As a result, the reviewed data mostly capture the full seasonal variability of C. thous fruit consumption patterns.

3.2 Frugivory

We recorded the consumption of 128 fruiting plant species in 81 genera and 43 families (Table 1). At least 18 species consumed by C. thous are exotic (cultivated). Overall, the average FO of fruits was 62.3% (±33.1; n = 29 studies), ranging from 9% to 120%. The highest frequency (120%) was recorded in a strictly protected conservation area, whereas the lowest frequency (9%) was observed in restinga (coastal tropical moist forests) and mangrove areas (Table S2). Two studies reported a total absence of fruits in the C. thous diet, specifically in a Brazilian Pampa biological reserve (length: 12 months) and in a Colombian Paramo (length: 4 days) (Table S2). Compared with other dietary items, fruits represented a substantial portion of C. thous diet overall (FO = 62.3% ± 33.1; Figure 2A). This pattern remained consistent across different habitats (native: FO = 68.8% ± 63.2, n = 13 studies; modified: FO = 63.1% ± 26.9, n = 15 studies; Figure 2B) and biomes (Atlantic Forest: 66.6% ± 37.6, n = 13 studies; Cerrado: 53.6% ± 29.8, n = 7 studies; Pampa: 56.4% ± 43.6, n = 4 studies; Figure 2C), highlighting the species' tendency to frugivory regardless of environmental context. The average PO of fruits was (34.6% ± 23.4%; n = 7 studies), from 9.6% to 75.01%. Fruit was the main item in the C. thous diet in 10 of the 29 studies (35.7%; Table S2). The mean number of fruit species consumed per study was 6.8 (±4.9).

Myrtaceae had the greatest species richness in the C. thous diet (25 spp. or 14.7% of the total number of species recorded), followed by Solanaceae (14 spp., 8.2%), Moraceae (14 spp., 8.2%), Anacardiaceae and Arecaceae (13 spp. each, 7.7%). At the species level, Syagrus romanzoffiana (Arecaceae) fruits were the most frequently consumed (16 studies, 43.2%), followed by Hovenia dulcis (Rhamnaceae—cultivated species; 10 studies, 27.0%), Psidium guajava (Myrtaceae—cultivated species; 7 studies, 18.9%) and Solanum lycocarpum (Solanaceae; 6 studies, 16.2%). The remaining fruit species were reported less consistently, with 85 of them being found in only a single study (Table 1).

3.3 Seed Dispersal Performance

The size of consumed fruits ranged considerably, from the small fruits of Myrsine sp. (4.0 mm width) to large ones like Artocarpus heterophyllus (295.8 mm width) (Figure 3A, Table 1). The mean fruit diameter was 30.8 ± 37.4 mm. Likewise, seed size from consumed fruits also varied greatly, from the small seeds of Piper aduncum (0.5 mm width) to the large seeds of Cocos nucifera (125.0 mm width) (Figure 3B, Table 1). The mean seed size was 7.6 ± 12.1 mm. However, the largest seeds defecated whole were of S. romanzoffiana (15.7 mm) and A. heterophyllus (17.4 mm) (Table 1). We also emphasise that fruits with particularly large seeds (e.g., C. nucifera, Persea americana, Mangifera indica) likely only had their pulp or endosperm consumed by C. thous (see Macdonald and Courtenay 1996; Rocha et al. 2004). For C. nucifera, its fruits likely need to be broken before (e.g., through gravity) C. thous can utilise them, as seen in the dispersal of other large-seeded palms (Brewer 2001).

Seeds ingested by C. thous are commonly defecated intact and viable (Table 2; Figure 4; Figure S2). Of the 32 fruiting plant species tested for germination after gut passage, 24 species germinated (75%; Figure 5). From those compared with a control test (n = 16 spp.), effects on germinability were positive in two species (12.5%), negative in four species (25%) and neutral in 10 (62.5%; Figure 5). We could not evaluate the effect on germination for nine species since they had no control comparison (Table 2; Figure 5).

For captive C. thous individuals, the average seed passage time was 7.7 h, measured for three sizes of artificial seeds and wild seeds from the fleshy-fruited species Sarcomphalus mistol, Vachellia aroma, Celtis tala and S. romanzoffiana (Table S2). C. thous defecates in all types of habitats but displays a general trend for open habitats, like shrub grasslands and wooded savannas (up to 94% of scat samples [n = 39] in Rio Pratudão Ranch, Central Brazil, Table S2). Scats are mostly found in unfavourable sites for germination, including trails, dirty roads and forest edges (n = 6 studies, Figure 4; RBC, pers. obs.). The canid can transport seeds distances of 8.9 to 10.9 km (n = 2 studies; Figure 4).

4.1 The Importance of Fruits in Cerdocyon thous Diet

Our results indicate that fruit consumption represents a substantial proportion of C. thous diet. Fruits were the main food item in 10 out of 27 studies (35.7%) that evaluated the C. thous diet. Fruits were also among the most common diet items with high FO (≥ 50%). Moreover, in some instances, fruit can account for up to 56.4% of the biomass intake (Juarez and Marinho-Filho 2002; for biomass percentage, see also Bueno and Motta-Junior 2004; Rocha 2008) and up to 44.0% of the volumetric intake (Facure and Monteiro-Filho 1996; for volume percentage, see also Macdonald and Courtenay 1996; Facure et al. 2003; Bossi et al. 2019). This is noteworthy considering the variety of biomes in which this canid's diet was evaluated—for example, tropical rainforests, savannas, dry forests, grasslands—and the wide diversity of fruit-bearing plants consumed by C. thous (Table 1; Tables S1 and S2; Figure S1). Although these biomes differ in overall vegetation structure and species composition, many of them offer a year-round supply of fleshy fruits (Kuester et al. 2020; Silva et al. 2020). This broad regional availability of fruit resources likely contributes to maintaining high levels of frugivory at the population level, as C. thous individuals across different habitats consistently incorporate fruits into their diet. While an individual's fruit consumption may vary depending on local availability, the overall pattern of frugivory remains high across the species' range due to the diversity and ubiquity of fruit-bearing plants in the Neotropics (Fleming et al. 1987).

As suggested by Vieira and Port (2007), FO of fruits in C. thous diet can be related to habitat type or the energetic demands of individuals. High fruit frequencies may be related to highly productive habitats in the Neotropical regions (e.g., forests) (Haugaasen and Peres 2007; Brockerhoff et al. 2017). Low frequencies may be associated with ecosystems where plants with fleshy fruits are less common (e.g., grasslands) or with the necessity of C. thous to rely upon high-energy food (e.g., small vertebrates) in colder regions (Vieira and Port 2007). This pattern is reflected in our results, where fruit FO tends to be higher in biomes with more complex vegetation structures (Atlantic Forest, Cerrado; Figure 2C) compared to more open habitats (Pampa; Figure 2C). Ultimately, particular behaviours of local C. thous populations (e.g., hunting efficiency) and limited sample size can also explain low frequencies of fruits in the species' diet (Pedó et al. 2006; Delgado-V and Zurc 2007).

Fruit-eating by C. thous varies seasonally, with debate over the dominant fruit-consuming season. Several studies indicate that fruit consumption is higher during the rainy season (Motta-Junior et al. 1994; Facure et al. 2003; Bueno and Motta-Junior 2004; Rocha et al. 2004; Vilagran 2004), which coincides with peak fruiting periods in many tropical rainforests and woodland savannas (Mendoza et al. 2017). Conversely, other studies suggest higher fruit consumption in the dry season (Vieira and Port 2007; Amaral 2007; Araujo 2008), potentially reflecting the phenology of plant species that fruit asynchronously or persist as dry-season resources, particularly in more seasonal environments like dry forests (Cortes-Flores et al. 2019; Pereira et al. 2022). There is also evidence to the effect that fruits may be consumed equally by C. thous throughout the year, irrespective of the season (Cheida 2002; Gatti et al. 2006a; Raíces and Bergallo 2010; Bianchi et al. 2014; Porto and Rui 2019), a pattern that may be associated with vegetation physiognomies where fruit availability remains more stable across seasons, such as wetter rainforests (Mendoza et al. 2017).

The studies encompassed a broad seasonal range, covering both wet and dry seasons, with most conducted along the central-south axis of Brazil (Table S2; Figure S1). This geographic concentration ensures robust temporal representation but may introduce regional biases that limit broader inferences across the entire Neotropical range. While our data set captures seasonal variation in C. thous fruit consumption, differences in fruiting phenology among biomes suggest that general trends may not be uniform across the species' distribution. Ultimately, C. thous' fruit consumption seems to be driven by the local fruiting patterns of the species exploited (Kuester et al. 2020), which may vary seasonally and geographically (Ting et al. 2008; Miranda et al. 2023). This pattern is unsurprising since C. thous has an opportunistic feeding strategy, consuming items that are readily available (Motta-Junior et al. 1994; Novaes et al. 2011; Santiago et al. 2023).

4.2 What Is on the Menu? Fruits Consumed by Cerdocyon thous

We found that C. thous feeds on at least 128 species of fruiting plants. Many of them (n = 85 spp.) had a single record of consumption, whereas only four species were frequently consumed. As noted by Facure et al. (2003), most fruit species consumed by C. thous have a low recurrence and contribute little to the total volume ingested; some, on the other hand, dominate and are consumed in great quantities (e.g., cultivated species). This fruit consumption pattern indicates a nonuniform resource use rather than a lack of dietary diversity (Jácomo et al. 2004). Instead of consuming all available food items uniformly, C. thous likely relies more frequently on a subset of resources, including specific fruit species, while incorporating others less frequently. We can only speculate about the reasons for that, but it may have to do with the differential availability of fruit species and/or with the canid's preferences guided by the nutrient contents of fruits. Thus, C. thous seems to concentrate its fruit consumption on a few species (Gatti et al. 2006b)—particularly S. romanzoffiana, H. dulcis, P. guajava and S. lycocarpum—highlighting this species' tendency to rely disproportionately on specific, key fruit items while maintaining an overall opportunistic feeding strategy.

In several studies, S. romanzoffiana was the most common fruit species in C. thous diet (Facure and Monteiro-Filho 1996; Rocha et al. 2004, 2008; Bossi et al. 2019), consumed nearly year-round (Facure et al. 2003; Kuester et al. 2020) and in such quantity that its faeces can be dominated by S. romanzoffiana seeds (Reitz 1974). S. romanzoffiana is a widely distributed palm tree in South America (Glassman 1987; Carvalho 2006; Lorenzi et al. 2010), commonly found in the wild and also used as an ornamental plant (Noblick 2017; Carvalho 2006; Soares 2023). This palm produces fruit year-round, with up to 800 yellow-orangish fruits per bunch (Galetti et al. 1992). These features likely explain its prevalence in C. thous diet.

C. thous consumes fruits of 18 cultivated species, two of which (H. dulcis and P. guajava) feature among the most frequently consumed. C. thous' consumption of cultivated fruits is well documented in the literature (Motta-Junior et al. 1994; Facure and Giaretta 1996; Facure and Monteiro-Filho 1996; Cheida 2002). Several studies in our literature review were performed in modified habitats, including rural, peri-urban and natural areas with different degrees of disturbance. The occurrence of C. thous in human-disturbed areas with high availability of cultivated plants (e.g., farms, orchards, crops, agricultural fields) may explain their high frequency on its diet (Facure and Monteiro-Filho 1996; Facure et al. 2003). Such is the case with the consumption of the Japanese raisin tree (H. dulcis; Rhamnaceae), one of the most important invasive species in Brazilian protected areas (Zenni and Ziller 2011; Hendges et al. 2012). H. dulcis is particularly common near human settlements and modified habitats, where C. thous is prone to consume it (Facure et al. 2003; Facure and Monteiro-Filho 1996; Rocha et al. 2008). The strong correlation between the abundance of H. dulcis and C. thous highlights the canid's role in spreading this invasive species (Lima et al. 2015). This raises concerns about C. thous spreading H. dulcis in natural habitats and protected areas (Amaral 2007), potentially interfering with native species' recolonisation due to H. dulcis' allelopathic properties (Boeni 2011). In fact, the phenomenon of canids acting as dispersers of exotic, potentially invasive plants is not uncommon, as most canids are opportunistic generalist feeders (Spennemann 2021).

4.3 Seed Dispersal by Cerdocyon thous