Foraging Behaviour of Brazilian Riverine and Coastal Fishers: How Much is Explained by the Optimal Foraging Theory?

Optimal Foraging Theory (OFT) is here applied to analyse the foraging behaviour of Brazilian artisanal ﬁ shers of the Atlantic coast (Itacuruçá and São Paulo Bagre villages) and of the inland Amazonian region (Jarauá and Ebenezer villages). Two OFT predictions are tested. Hypotheis1: A ﬁ sher who travels to more distant sites should return with more ﬁ sh, and Hypothesis 2: The further a ﬁ sher goes, the longer s/he should stay ﬁ shing in a patch. OFT did not explain ﬁ shers’ behaviour (non-signi ﬁ cant regressions for coastal villages) or explain it in speci ﬁ c seasons (low water season for one Amazonian village: H1 r 2 =24.1; H2 r 2 =37.2) and in speci ﬁ c habitats (e.g., lakes and backwaters in Jarauá village, Lakes: H1 r 2 =13.5; H2 r 2 =24.0; Backwaters: H1 r 2 =34.4; H2 r 2 =46.5). The ﬁ ndings can indicate areas or seasons that are under higher ﬁ shing pressure, when ﬁ shers try to get the best out of a situation without any concern about resource conservation. By knowing the variables that in ﬂ uence ﬁ shers’ decision-making processes, management initiatives may be more ﬁ ne-tuned to the local reality and are thus more likely to succeed.


INTRODUCTION
Originating in microeconomic theory (Rapport & Turner 1977), optimal models were fi rst applied to understand animal foraging behaviour (Stephens & Krebs 1986). Their potential to explain human foraging behaviour through simple, operational and realistic approaches was soon realised (Winterhalder 1981). Such models offer plausible explanations for a variety of questions that goes from human settlement pattern to the size and composition of social groupings (Winterhalder & Smith 1981), and have been useful in estimating foraging behaviour in archaeological studies (Bettinger 1991). The basic assumption of optimality theories states that the foragers' decisions aim at the maximisation of their fi tness (Stephens & Krebs 1986). However, measuring fi tness in humans is usually a hard and even unfeasible task due to practical and ethical reasons. One of the alternatives is to choose a short-term energy-return currency, assumed to have direct implications in fi tness, as fi tness is supposed to be positively related to the rate of energy intake acquired while foraging (Winterhalder Foraging behaviour of Brazilian fi shers / 237 1981). Successful examples using such indirect measures can be found in archaeological studies (Bettinger 1991;Burger et al. 2005;Lupo & Schmitt 2005) as well as in studies among contemporary indigenous groups (Winterhalder & Smith 1981;Smith & Winterhalder 1992;Bird & Bliege Bird 1997).
The fi rst optimal foraging space-use model-the patch choice model-was developed by MacArthur and Pianka (1966) and explores the selection of foraging areas in a heterogeneous environment. Charnov (1976) developed another model that dealt with the selection of a foraging pathway, the marginal value theorem, predicting when a forager should leave a patch. This model assumes that the foraging activity in a patch reduces the food availability in its immediate vicinity. An optimal forager should leave a patch when the marginal intake rate in that patch drops to the average rate of intake in the overall habitat (Charnov 1976). Orians and Pearson (1979) developed yet another model-the central place model-which can be understood as a variation of the marginal value theorem, where the forager has a central place (a house, a village, etc.) to return to after foraging. Doing so, an optimal forager should maximise the rate of energy delivered to the central place, including the expenses involved in the round trip to the foraging ground. This model has been widely applied to human foragers, as they usually have fi xed settlements (central places) where they return after a foraging day (Glover 2009).
Optimal foraging models, such as central place foraging, show potential to go beyond the understanding of the evolution of human behaviour. They offer an alternative to study local resource management by demonstrating whether foragers forego short-term benefi ts for long-term ones through sustainable harvests (Alvard 1993;Hames 1987). A constant maximisation strategy, as assumed by optimal foraging, implies that foragers would not refrain from overexploiting their resources, if needed. Nevertheless, it is demonstrated that, depending on the biological characteristics of the species (e.g., annual maximum sustainable yield, intrinsic rate of increase), foragers can behave according to predictions of the optimal foraging theory and still exploit their local resources in a sustainable way (Alvard 1998). Sustainability in this case is not a synonym for conservation, the latter being a side effect. Aswani (1998) was one of the fi rst persons to apply the foraging theory to understand marine resource management strategies. Studying fi shers from the Solomon Islands, he showed that it is possible to integrate optimisation model studies to provide practical management suggestions for the sustainability of long-term fi sheries. In Brazil, optimal foraging studies have also been applied to investigate fi shers' behaviour in freshwater and marine environments (Begossi 1992;Begossi & Richerson 1992;Begossi et al. 2005).
In our study, using examples of artisanal fi shers from the Brazilian Amazonian region and southeastern Atlantic coast, we addressed the central place foraging model by examining two of its main predictions (hypotheses): a) fi shers should stay longer when foraging in more distant spots, and b) by doing so, they should catch more fi sh in those spots. By considering different case studies and using OFT as a tool, we wanted to better understand which factors (gear, seasonality, types of habitat exploited, etc.) are relevant to defi ne the behavioural strategy adopted by fi shers in different places. The results also have management implications: If fi shers are predominantly 'catch maximisers', such behaviour must be carefully considered when developing management strategies, as they will invest in short-term benefi ts, trying to get the best out of a situation, which can mean exploiting fi sh stocks to the limits, regardless of the resource's abundance.

Study Sites
Four sites-two on the Atlantic coast (Itacuruçá island and São Paulo Bagre) and two in the inland Amazonian region (Jarauá and Ebenezer)-were studied, as they represent small villages where fi shing is the main subsistence and economic activity ( Figure 1).
The research in Itacuruçá island, Rio de Janeiro State, was conducted at fish-landing points at Itacuruçá beach (22 o 55'92"S and 43 o 54'83"W) and in Gamboa (22 o 55'90"S and 43 o 53'73"W), from September 1989 to February 1990. During this period, there were 26 families living there, all dependent on fi sheries to some degree. Local fi shing is mostly practised by men, although children and women do fi sh for subsistence at times. Besides fi shing, the people also work as maids or housekeepers for tourists. Local fi shers generally use gillnets and encircling nets to catch mainly whitemouth croaker (Micropogonias furnieri, Sciaenidae), rays and catfi sh (Ariidae), besides shrimp (Penaeidae) (Begossi, 1992(Begossi, , 1995. São Paulo Bagre is a fi shing community located in the estuary of Iguape-Ilha Comprida (24°57'51''S and 47°53'13W''). Shrimp fi shing (Litopenaeus schmitti, Penaeidae) is the main activity, carried out by an artisanal fi shing method using the 'gerival', a small mesh-sized gillnet attached to a pole and trawled on the bottom of the estuary by a fi sher in a paddled canoe (Hanazaki et al. 2007). The São Paulo Bagre community comprises 17 families, who base their economy on fi sheries, shrimp collection to be sold as bait, subsistence agriculture and tourism-related activities, such as working as boat captains for tourists. Plants and timber are rarely extracted (Hanazaki et al. 2007). Here, fi shing is mostly by men. Jarauá (02 o 51'849 S, 64 o 55'750 W, Amazonas State) is a fi shing village located at the confl uence of the Japurá and Jarauá rivers, in the Mamirauá Sustainable Development Reserve, home to 35 families in 1994 (Queiroz 1999), but this fi gure has probably changed signifi cantly in the last decade. The Mamirauá Sustainable Development Reserve has been under management since 1990 with special focus on two important commercial fi shes-tambaqui (Colossoma macropomum, Serrasalmidae) and pirarucu (Arapaima gigas, Osteoglossidae) (Queiroz & Crampton 1999;MacCord et al. 2007;Castello et al. 2009;Silvano et al. 2009). Their comanagement programme involves rotation of fi shing pressure among lakes, monitoring (counting) of pirarucu by local fi shers and participation of fi shers and other community members in decision-making processes, including the enforcement of the rules. Because of its successful local management process, the Mamirauá Reserve was recommended for the international fi shing certifi cation awarded to products that come from well-managed and sustainable fi sheries by the Marine Stewardship Council (Wilson et al. 2001). This is true of Mamirauá as a whole, and Jarauá specifi cally, where formal co-management has been successfully carried out in the last 20 years, resulting in a signifi cant increase in some fi sh populations (e.g., pirarucu, tambaqui), as well as in socioeconomic improvements to fi shers (higher income), despite an increase in the number of fi shers entering the fi shery (Castello et al. 2009;Silvano et al. 2009 (Viana et al. 2004). Tambaqui and pirarucu are also important commercial fi shes in Ebenezer, but the different geographic and social features here result in a greater dependence on migratory fi shes, such as many catfi sh species (MacCord et al. 2007). Both Jarauá and Ebenezer villages practise slash-and-burn agriculture, mostly focused on cassava to produce cassava fl our. Although fi shing is the main economic activity, fi shers and their families can be involved in, and get paid to, work in the Reserve projects, such as forest, caiman, turtle or bird management. In both communities, children and women fi sh mostly for subsistence, although wives sometimes help their husbands on longer fi shing trips.

The Habitat Types
The habitats regularly exploit ed by Jarauá and Ebenezer fi shers are: • River: The main river channel, which is larger and deeper than other habitats and with a faster water fl ow. Fishers usually exploit the river to catch large migratory fi shes, such as catfi sh (Pimelodidae). • Flooded forest: Also locally called 'várzeas', this ecosystem is created during the high water season, when water from the main river and lakes fl ood the adjoining forest, forming an important environment for fi sh feeding and nursery grounds. • Lakes: These floodplain lakes are usually seasonally connected to the main river and to one another during the high water season, when they form lake systems. Some of the largest lakes in the Mamirauá Reserve were created by a channel that was separated from the main river due to sedimentation. Mostly whitemouth croaker (Micropogonias furnieri, Sciaenidae), catfi shes (Ariidae) and ray, b Litopenaeus schmitti, Penaeidae, c small mesh-sized gillnet attached to a pole and trawled on the bottom of the estuary by a fi sher in a paddled canoe • Connecting canals: Locally called 'paranás', these channels link the fl oodplain lakes to the river or link together several floodplain lakes during the high water season. These channels usually increase considerably in size during the fl oods, but become much shallower or even dry out during the low water season. • Backwaters: These are lakes that are permanently connected to the main river (or its tributaries) by an openended mouth. These backwaters are thus more accessible. Water characteristics and fi sh assemblages may experience less seasonal change than lakes, which are not linked to the river during the low water season. More details about aquatic habitats and fi shing communities of Mamirauá can be found in other surveys (Crampton 1999;Henderson & Robertson 1999;Silvano et al. 2009).

Procedures
Data from fishing trips were assessed at landing points, gathered directly from fi shers. A total of 113 fi shing trips were sampled monthly (six consecutive months) in the spring and summer of 1989-90 at Itacuruçá island. Monthly appraisals of fi shing trips and catches were collected in 1999-2000 in Sao Paulo Bagre for 10 consecutive months, totalling 204 fi shing trips. In the Amazonian communities, fi sh landings were assessed during the high water (17 consecutive days in June) and low water (15 consecutive days in October) seasons, as logistics made monthly evaluation diffi cult. This resulted in 268 fi shing trips in Jarauá and 204 in Ebenezer.
Fishers were asked about the distance (in minutes) travelled to the fi shing spot for each trip, the time spent fi shing, fi shing gear used and fi sh composition. Fish caught were weighed by the researchers and identifi ed to the nearest possible taxonomic level. More detailed data about these four fi shing villages are available in other surveys (Begossi 2006a;Hanazaki et al. 2007;MacCord et al. 2007). Simple linear regressions were used to analyse the data after normalisation through natural logarithm when needed. It is assumed, based on interviews, that in most of the fi shing trips fi shers visit only one fi shing spot. We performed two separate regressions to answer the questions below: 1. Do fi shers catch more fi sh when they forage in distant patches? Dependent variable (y): amount of fi sh caught (kg) X independent variable (x): travel time (min).
2. Do fi shers stay longer in more distant patches? Supposing there is no resource depletion and that the resource is evenly distributed in the environment, fi shers will have to stay longer in the patch (fi shing spot) to catch more fi sh. Dependent variable (y): time fi shing (min) X independent variable (x): travel time (min).
If Regression 1 is signifi cant but not Regression 2, there can be some evidence of resource depletion or unequal resource distribution, which was not foreseen before the samplings, as the projects developed at each of the studied sites have different goals. Finally, as fi shers in the Amazonian region use a diversifi ed set of aquatic habitats (river canals, lakes and backwaters) and seasonality is clearly defi ned by the level of water (six months of fl ooded forest), it is likely that such factors can affect the predictive power of the models used. To test this, data about Amazonian fi sheries were also analysed separately by environment exploited and season. This is not expected to be a problem on the coast, where seasons are less defi ned and both the villages exploit just one kind of environment each (São Paulo Bagre: estuary; Itacuruçá : open ocean). Table 1 summarises the main features of each fi shing village and the average values for the variables used in the regressions. The fi shing communities showed differences not only between the two main environments (Amazonian region and the Atlantic coast), but also between fi shing communities in the same environment, as within the Amazonian region (Table 1). For example, the time fi shers spent fi shing is much longer in the Amazonian region (average=534 min) than on the coast (average=243 min), while the travel time varied according to the village, regardless of the environment. On the other hand, in a coastal village whose main resource is shrimp (São Paulo Bagre), fi shers do not fi sh for too long (average=148 min) and fi sh close to home (average travel time=29 min), while still assuring the highest returns in weight of catch (Table 1).

Applying the Models to Field Data
The fi rst hypothesis proposed in this study and one of the core questions in the central place foraging model states [Downloaded free from http://www.conservationandsociety.org on Friday, January 13, 2012, IP: 129.79.203.177] || Click here to download free Android application f this journal that a forager who travels further must bring back home a higher energetic return than when foraging closer to the central place. This assumption is reasonably confirmed by most of the linear regressions (Table 2, r 2  SPBagre =22%,  r 2 Itacuruçá =19%, r 2 Ebenezer =12%; P<0.001). The only exception is one of the Amazonian region villages, Jarauá (P>0.05). The other Amazonian region village, Ebenezer, had a signifi cant regression coeffi cient, but it showed that only 12% of the return in kilograms of fi sh caught was explained by the distance travelled by the fi shers (Table 2).
If the result demonstrated above was the consequence of an optimal strategy by fi shers, then it is expected that the fi shers stayed longer in the more distant patches in order to catch more fi sh. However, this is not consistently observed among the villages. One of the coastal villages (São Paulo Bagre) had a signifi cant but very low regression coeffi cient (Travel time barely infl uenced Time fi shing; r 2 =9%), while the other coastal village (Itacuruçá island) did not show a signifi cant result (P>0.05), despite the signifi cant results in the fi rst regression based on the central place foraging model (Table 3). The Amazonian region villages were even more surprising: Jarauá, which was not signifi cant in the fi rst regression, now presented a very high regression coeffi cient (r 2 =43%). If a Jarauá fi sher goes to a place further away, he stays longer there. But in Ebenezer, the other village, the distance of a fi shing spot did not affect the time spent there by a fi sher (P>0.05) ( Table 3).

Foraging Models and the Specifi cities of Each Situation
The fi rst questions that might arise in observing the results in Tables 2 and 3 will be about the variables that infl uenced the fi shers' decision on how far to go fi shing. What are they optimising: Total catch (quantity) or a selective catch of a given fi sh species? Which variables are relevant: Gear used, fi sh target, season or habitat? In the case studies presented, fi shers from only one village, the coastal São Paulo Bagre, used one single kind of gear to catch shrimp, while the others used a mix (Table 1). The same two regressions performed with the whole data set are shown for Itacuruçá island, but now separated by gear (Table 4). Confi rming what was observed earlier, the results show that increasing the distance of the fi shing spot from the central place increased the amount of fi sh caught, especially using a particular kind of gear, such as the encircling net ("rede de aperto") (r 2 =43%; P>0.001), but this does not imply that fi shers will stay longer in more distant patches (Table 4).
For the Amazonian region villages, although the gear used varied consistently according to the seasons, we preferred to analyse the data by 'season' instead of by 'gear' because we believe that the gear is chosen in conformity to the local water level. Apparently, during the low water season, Amazonian fi shers behaved as predicted for optimal foragers (Ebenezer r 2 =25%; Jarauá r 2 =24%; P>0.01), but this was not always the case during the high water season (Table 5). There is yet another factor to be considered in these villages-the different habitats exploited. When the two regressions were performed by habitats, the results were highly variable. The distance travelled by the fi sher explained both the amount of fi sh caught and the time the fi sher remained in a patch (as expected by the model) for some habitats only ( Table 6). The only consistent result for both villages was that the model explained fi shers' behaviour when fi shing in lakes or in backwaters, which are similar to lakes (Table 6). It is worth noticing that for these habitats, travel time explained most of the variability in the amount of fi sh caught and in the time spent fi shing in a patch (Table 6).

DISCUSSION
The application of optimal foraging models to understand human behaviour reached its peak in the middle of the 1980s, following important studies done among the Inuit and the Cree in Canada (Smith 1981;Winterhalder 1981) and among South American groups and Australian aborigines (O'Connell and Hawkes 1981;Hawkes & O'Connell 1982;Beckerman 1983).
After this period, a few studies were published concerning human foragers, most of these focusing on archaeological applications, including the development of new models (Metcalfe & Barlow 1992). Only by the end of the 1990s was there a resurgence of optimal foraging in anthropology and archaeology (Zeanah 2002;Lyman 2003;Lupo & Schmitt 2005), and for the fi rst time fi shers and shellfi sh gatherers came to be the focus of such studies (Begossi 1992;Bird & Bliege Bird 1997;Aswani 1998;de Boer et al. 2002;Thomas 2007). Until then, hunters and gatherers were the only ones to be considered in human ecological studies, apparently because such human groups represented the last 'real foragers'.
With these recent studies, it has become clear that the study of fi shers and shellfi sh gatherers could also bring insights to the understanding of the evolution of human behaviour (Bird & O'Connell 2006;Nordi et al. 2009), of archaeological facts, such as the formation of shell assemblages (Bird & Bliege Bird 1997;Thomas 2007), and it could even help establish management measures or understand local tenure systems (Aswani 1998). These more recent applications of optimal foraging models also show that new and more elaborated  models are not necessarily required to understand 'current foragers', whose concerns are not only to eat and bring provisions to their families, but also to sell their foraging product (e.g., fi sh) to buy other goods and foodstuff (Begossi & Richerson 1992). Advocates of simple models have shown that they can be widely used in different contexts with satisfactory outcomes, although sometimes it is necessary to reformulate the hypotheses and reconsider the currencies used (e.g., calories, protein, money) (Bird & O'Connell 2006). This is what we did in this study: We showed how a simple and well established model (central place model) can explain different fi shers' behaviour in the exploitation of different resources, considering the peculiarities of each region. In our study, when we used the whole data set without considering differences in gear, habitats or seasons, we observed that the fi rst hypothesis (s/he who travels further catches more fi sh) was confi rmed for some villages, especially the ones on the coast. However, the second hypothesis, which should be viewed as a consequence of the fi rst one in the original model (the further a fi sher goes, the more time s/he should spend in the spot), was not confi rmed for most of the villages. The results were diverse and no pattern could be established either in the Amazonian region or at the coast.

Table 6 Simple linear regressions for the Amazonian communities analysed by habitats, having 'travel time' as the independent variable. Data was transformed in natural logarithm. Only habitats with more than 15 fi sh landings were analysed in order to avoid biased results due to small number of observations
The model seems to work well when no or a minimum amount of variation is included in the foraging activity. For example, the model showed a good fi t for the only situationthe coastal São Paulo Bagre-where fi shers used one kind of gear (gerival), exploited one kind of habitat (estuary) and one kind of resource (shrimp), and where there are no pronounced seasonal differences.
In places where the fi rst hypothesis but not the second one was confi rmed (e.g., the coastal Itacuruçá), one can initially surmise that local fi shers, instead of optimising, are dealing with local resource depletion; they need to go further to bring some food home, because nearby fi shing spots were already depleted. Alternatively, time fi shing in more distant fi shing spots (or total fi shing time including travel and time on the spot) may be limited by the available ice (important to avoid fi sh or shrimp spoilage), as observed for Amazonian fi shers at the Negro river . This would be an example of a constraint not included in the original model.
Fishing is a complex and unpredictable activity, affected by external factors such as the weather, and by individual choices, such as the use of specifi c gear. Fishing can be performed in different ways and with the use of an array of different methods and types of gear. Some of these require more effort and the constant presence of the fi sher (e.g., diving, hook and line), while others do not (e.g., set gillnets), even allowing the performance of parallel activities (e.g., a fi sher can set the nets and go back home until s/he decides to check the nets). If this is the case, it is reasonable to suppose that the gear used or the season will affect the fi sher's behaviour. Such factors might represent important variables taken into account by fi shers when they make their decisions about where, when and how long to stay fi shing.
When we considered these factors, we did observe patterns.
Kinds of gear, for example, are an essential variable to explain the amount of fi sh caught in relation to the distance travelled for one of the coastal villages, Itacuruçá. For some gear, such as encircling and entangling nets, the distance travelled explained the catch, while this is not true for set gillnets. For none of these gears, however, going further implied staying longer. These results indicate a 'non-optimal' behaviour for these fi shers, as it was not explained by the foraging models. Non-optimal behaviour may occur due to several reasons: For example, an open access situation where a fi sher tries to get the most of a fi shing spot before others arrive to exploit the same spot (Begossi 1992), or a time lag between environmental changes and the selection of optimal behaviour (adaptive lag) (Laland & Brown 2002: 142; but see Laland & Brown 2006 for a different opinion on the lag between environmental changes and human adaptation). Glover (2009) showed that even a false public announcement of good foraging places can induce non-optimal behaviour, such as human foragers staying longer in non-productive patches. Seasonality also showed to be an important variable taken into account by fi shers in their foraging activity, depending on the situation. In the Amazonian region villages, we saw that in one of them (Jarauá), fi shers optimised in the low water season. In this season and specifi cally in this village, fi shers focus on the capture of pirarucu, a large fi sh that occurs in lakes and comes to the surface every 5 to 15 minutes to breathe. Pirarucu fisheries are regulated by a co-management agreement, occurring in a very specifi c period of the year and following a quota system defi ned yearly (Castello et al. 2009). During the few days when pirarucu fi shing is allowed, fi shers have to do their best to reach their quota and it is not surprising that they optimise their fi shing behaviour, as predicted by the models.
Finally, the third variable we considered-the kind of habitat exploited-was also important to the degree to which foraging models can predict fi sher behaviour. Again, our results indicated that fi shers' optimisation as predicted by models was only reached in some habitats, namely lakes and backwaters. Both habitats have well-defi ned boundaries, even though both can be connected regularly or temporarily to the main channel. During the low water season, the prey (fi sh) has usually nowhere to go. Lakes are thus very similar to a patch as idealised in the original model proposed by Charnov (1976). Having conditions more similar to the original model can indeed assure a better fi t, but, as we showed before, this is not necessarily required. Aswani (1998), studying the fi shers from the Solomon Islands, also considered different habitats when analysing optimal behaviour, but in this case each major habitat category considered was a large patch. Aswani confi rmed the predictions of the optimal foraging models for the studied fi shers, showing that fi shers spend more time in the more productive habitats (patches) and also spend more time in the fi shing spots of less productive habitats in a given season (Aswani 1998).
The analyses made in this and in other studies (Aswani 1998;Thomas 2007;Glover 2009) were performed at the population level, therefore not addressing individual behavioural variation.
Nevertheless, individual variation can affect the observation of an optimal behaviour. Individuals vary in several aspects: Ability to learn, which also depends on life stage (Bird & Bliege Bird 2000); need to acquire information before they can behave optimally (Clark & Mangel 1984); and variations in proneness to take risks (Smith & Wilen 2005). Such factors are important approaches to a fi ne-grained analysis, and could be investigated in the future, perhaps even including the fi shers' own explanation for their behaviour. Nevertheless, a population level of analysis may be the most useful one to support fi shery management approaches (Aswani 1998), as it shows the behaviour of the majority of individuals that will ultimately affect the exploited resources. Table 7 summarises the main results of published Brazilian studies on fi shers' optimal foraging. In general, these studies approach at least one of the two hypotheses tested here: A fi sher should catch more fi sh if s/he travels further or a fi sher should stay longer in a spot if s/he travels further. Some of these studies also indirectly test a third hypothesis: A forager should leave the patch at the optimal time, after which the costs of searching for the prey will be higher than the potential benefi ts of catching additional prey. This assumption is hard to test without measuring the forager's load curve, which can be done experimentally. It is only possible to provide indirect evidence of this hypothesis if it is shown that there is a negative correlation between the time spent fi shing and the amount of fi sh caught: Fishers should stay for shorter time in productive patches.

Comparing Studies on Foraging Models Applied to Brazilian Fishers
The nine studies found in the literature seem to show a good fi t to the model because they considered, with a few exceptions, different gear and seasons separately, but the results for the whole data set have not been presented (Table 7). Even though this is just indirect evidence, such studies seem to confi rm the relevance of including pertinent variables that describe the environment (e.g., seasons or habitats) or foraging methods (e.g., types of gear) in the optimal models as a means of understanding human foraging behaviour.
Another important point interpreted from these published studies is that focusing on one prey, especially on a less mobile prey (such as shrimp), is key to a better fi t of foraging models to fi shers' behaviour (optimisation). Fishers can more easily access the location and density of a less mobile prey in a foraging patch. This is suggested by Begossi (1992) when explaining the optimal behaviour for Sepetiba shrimp fi shers on the southeastern Brazilian coast.

CONCLUDING REMARKS: OPTIMAL FORAGING AND MANAGEMENT
The results from our survey and the comparison with other published studies show that simple optimal foraging models can be more widely and successfully applied to current foragers, if the details and particularities of each situation are taken into account, such as changes in behaviour due to seasons, climate change, foraging gear, and habitat variability. By doing so, we can be more convincing when applying foraging models to predict fi shers' behaviour and thus state whether specifi c foragers do or do not optimise. More than that, the fact that a forager optimises in one situation does not imply s/he will always do it. As shown here, seasonal changes may affect the perception a forager has of the pursued resource, new kinds of gear are introduced and it takes time to learn how to get the best out of them, and different environments may offer additional diffi culties in their exploitation. This has important implications for resource management, especially regarding new management approaches where resource users are an important part of the measures adopted (Warner 1997). In these approaches the way users behave are considered in the rules developed, which are sometimes proposed and discussed by the users themselves (Castello et al. 2009). Optimal foraging can then bring insights about when, where and why users go fi shing and to what extent they are optimising their fi shing returns. Management can refl ect more accurately on the reality if we know the variables that infl uence fi shers' decision-making processes (Béné & Tewfi k 2001). In the case of fi sheries, optimal foraging can help decide, for example, which areas or periods are under high fi shing pressure (i.e., when fi shers optimise regardless of the resource status, maximising their short-term harvesting rate). Alvard (1993) was one of the fi rst to show that groups that depend directly on natural resources may not be averse to overexploitation. Despite that, sustainable harvests can still happen and conservation would be, in this case, a side effect (Alvard 1995), depending on the size of the human group exploiting the resources, the biological characteristics of the prey, imperfect information about the environment and withingroup rules to control resource exploitation, among others. In our study, regardless of the reasons that explain optimal or non-optimal behaviour, there are periods or methods that can potentially put the resource under high exploitation pressure, while others work as release phases when the fi sh resources could potentially recover. For example, in the Amazonian region, slash-and-burn cassava agriculture is essential for fi shers' subsistence, as cassava fl our represents their main source of carbohydrate. They have to share their working time between fi shing and planting/harvesting, which may reduce fi shing pressure (Silva & Begossi 2009). This was also the case on the Atlantic coast, but continuous regulation by the Brazilian Federal Environmental Agency hampered the slash-and-burn agriculture, which could have increased the fi shing pressure. We cannot be sure if that indeed happened, but apparently coastal fi shers shifted from agriculture to tourism-related activities as well (Begossi 2006b).
In general, Brazilian fishers, especially those in the Amazonian region, do not seem concerned with conservation per se, but with assuring a steady or increasing use of resources, even if they have to regulate their fi shing activities to guarantee future use of resources. This pattern reinforces what the optimisation results have shown here. This is the case of the Amazonian fi shing agreements, where lakes are offi cially closed to outsiders or have their access controlled by local artisanal fi shers. In return, fi shers have to regulate their own exploitation as well, which can be done by gear regulation, quotas, seasonal access and even a mix of different measures, which are re-evaluated after a certain period (usually from three to fi ve years) (Lopes et al. 2011). By regulating their exploitation fi shers can achieve unintentional conservation, as observed also in the Pacifi c islands (Aswani 1998).
Fisheries management measures, in this case, could use the knowledge obtained from optimal foraging to establish access rules using the observed patterns of behaviour. Some of the fi sheries of this study, such as Jarauá and Ebenezer, are well-managed fi sheries that embody monitoring processes and adaptive co-management (MacCord et al. 2007;Castello et al. 2009). For them and based on what we found here, more specifi c measures could be adopted to protect the fi sh resources during the high water season, when fi shers are already nonoptimising and would probably be more willing to accept new regulations. The same idea could be applied for any other village, based on what, when and how fi shers are optimising or not optimising their exploitation.
Different fish resources will support different fishing pressures and this also needs to be taken into account. If the biological information of the species exploited is associated to the fi sher's behavioural information, better management strategies can be delineated and applied not only in Brazil and not only to artisanal fi shers (Bergmann et al. 2004;. Simple models, as the one applied here, have shown to be robust enough to help understand human behaviour concerning the use of different resources and at different levels of exploitation.