Two species of mammal, R. flavipectus and S. murinus, were captured in the households of villages endemic for plague in Yunnan and the predictor profiles of each were somewhat different. Both species are known to be reservoirs for plague [14, 15] and have been previously reported from southern China, including south-west Yunnan and the coastal areas of Zhejiang, Guangdong, Guangxi, Fujian, Hunan and Taiwan [16–18]. R. flavipectus, belonging to the order Rodentia, family Muridae, has been reported to be the main animal host and infectious source of plague in the commensal rodent plague foci in this region  and has been shown in surveillance data to be the dominant small mammal in Dehong prefecture, accounting for about 74% of small mammals captured , similar to the 80% in the current study.
S. murinus, belonging to the order Soricomorpha, family Soricidae, is reported to be an important reservoir of plague in Vietnam and Myanmar [18, 21]. This species accounted for 19% of small mammals captured in earlier surveillance data, similar to the 20% in the current study.
Other species of small mammal have been occasionally trapped in surveillance operations, including Mus caroli and Mus musculus (order Rodentia, family Muridae), were not found in the current study, probably due to the small sample size.
Although previous studies have reported an inverse association between the abundances of R. flavipectus and S. murinus in Yunnan and Guangdong Province of China [17, 22], no such association was evident in the current study. This may suggest somewhat different ecological niches in the current setting. This would be compatible with the somewhat differing predictor profiles of the two species. However, the small numbers of S. murinus captured resulted in low power to detect weak predictors of this species abundance, and this may account for some of the apparent difference in predictor profiles.
In any attempt to interpret the predictors of small mammals captured in traps it must be borne in mind that in order to capture an animal, it must first be present in, or pass by, the vicinity of the trap. It also must be susceptible to the lure of the bait, not afraid of the trap and not wise to the deception. In this study, traps and bait which have proved successful in routine capturing of small mammals in the same setting were used. However, it is likely that the effective lure that a bait presents depends not only on the nature of the bait itself but also on the state of hunger of the animal. Thus, predictors of mammal capture may relate to the presence of mammals in the house or to the likelihood of a mammal being attracted to the bait and succumbing to the deception. This may be the explanation of why more animals were caught in houses where food was stored in metal drums than in those where food was stored in sacks – the former presumably being considerably more small-mammal proof than the latter and therefore offering a less plentiful food supply, leading to a higher hunger level.
One of the most notable differences in the predictor profile between the two species is in the magnitude of the relationship between type of food storage and animal abundance. While storage of food in metal drums rather than in sacks appeared to favour the capture of both species, the effect was over four times greater in the case of S. murinus. Furthermore, other forms of storage – most commonly covered baked- earth enclosures – were associated with capturing fewer R. flavipectus but more S. murinus. This situation would be compatible with a situation in which the enclosures were penetrable by R. flavipectus but not by S. murinus. Given the differences in dentition of these two species – R. flavipectus adapted for gnawing and S. murinus for a mainly insectivorous diet [23–25] – this explanation has some biological plausibility.
The lower numbers of S. murinus caught in households around which vegetables were grown may also be an effect of the additional food supply for this species, which is reported to be an opportunistic feeder and whose diet includes plant material in addition to a wide variety of invertebrates and human food items . Similarly, R. flavipectus, whose habitat is reported to include garbage dumps , may find a ready supply of food in garbage around the household as well as in maize plantations in the village and therefore have less interest in taking trap bait.
That these predictors of mammal capture act via the susceptibility to capture is also supported by the lack of discernable association with the sighting of small mammals in the household, a variable that is expected to be more closely representative of the presence of small mammals in the house.
Other aspects of the physical surroundings identified as being associated with higher numbers of R. flavipectus caught were location of the household adjacent to other houses and, at the village level, the presence of communal latrines. As R. flavipectus is a commensal rodent, an area of relatively dense housing might be expected to support a larger population of rats than an isolated house, and therefore the presence of adjacent housing is more likely to act via an influence on the numbers of rats rather than on susceptibility to capture. Relatively high levels of infestation have been reported in areas of high housing density with more than 500 dwellings in the immediate vicinity  and it was postulated that a nearby house might act as a source of rodent infestation, especially as the range of rodents may well encompass more than one house at a time and dispersal is more likely to be successful over a short distance. The home range and movements of small mammals in rural areas have been reported to be wider than in the urban habitat [29, 30]. Our finding that location of the household close to other houses was associated with a lower probability of seeing small mammals, however, does not seem to be consistent with a higher level of infestation. Further investigation may be necessary to clarify this apparent anomaly.
The abundance of R. flavipectus was increased in villages with communal latrines. In general, latrines in rural areas are often semi-open and the sanitation is poor. This environment can offer shelter for rats and promote infestation. However, there seems to be some contradiction with the finding that households having an outside toilet had fewer rat captures. Further study of rat behaviour may be needed to better understand these relationships.
A cat in the house was associated with fewer rat captures and dogs in the house with greater numbers of rat captures. It is commonly understood that cats are important in the control of rats [28, 31]. Sixty percent of households either kept cats or allowed cats from the neighbourhood to freely enter and sleep in the house. These households were also less likely to report seeing small mammals in the house, in accord with the finding of a previous study in southern Africa in which families that kept cats had a lower likelihood of seeing rodents than those without cats . That study concluded that one cat could remove about 28 rats per year . However, Childs (1986) has postulated that cats are ineffective in eradicating existing rat infestation because they kill mainly young rats . Other studies have revealed that cats consume very few rats if food resources for the cats are plentiful . Another study even found a positive relationship between the presence of cats and rats in urban areas, which was attributed to the common benefit – waste food . There was no evidence that the presence of cats in the household affected the capture of S. murinus. It may be that the strong odour secreted by the well-developed scent gland of S. murinus deters potential predators .
Raising guard dogs is quite common in rural areas of Yunnan  and about one third of households raised at least one guard dog. Dogs were generally allowed to enter only the dining area and only in the daytime, in contrast to cats, that were allowed to roam freely throughout the house unchecked, and at any time. This difference may partially explain why the number of small mammals was not reduced by keeping a dog. In fact, the number of small mammals caught in households with a dog was increased. A similar finding was reported in urban areas of the USA where companion animals, especially dogs, were associated with the presence of rats because of the food and shelter provided by the pet owner .
Despite the identification of several predictors of R. flavipectus capture, other, as yet unidentified, lifestyle differences between the Dai and other, principally Han, ethnic groups, may also influence rat capture. Additionally, the reason for a higher probability of sighting of small mammals in households that keep ducks is not understood.
As discussed, numbers of animals trapped may reflect not only mammal population size but also susceptibility to capture, and trap setting for only two nights almost certainly underestimates the population size. The number of S. murinus captures was small and thus may not be powerful enough to detect some important predictors. Drawing conclusions regarding small mammal abundance from trapping results may need to be validated with other survey techniques. Furthermore, this cross-sectional study may be influenced by season and the results may change over the year.