Researchers who study habitat selection have proposed various models for the process. Marine biologist Peter F. Sale hypothesized the existence of a simple mechanism of habitat selection in fish that is based on levels of exploratory behavior. Sense organs monitor specific stimuli in the environment and send a summation of pertinent stimuli back to central-nervous-system centers, which regulate the amount of exploration. As the constellation of cues approaches some optimum level, exploratory behavior ceases and the animal stays where it is. An alternative hypothesis is that an animal has a cognitive map of the ideal habitat and that its behavior is goal directed. However, working with a species of surgeonfish, Sale tested juveniles in laboratory tanks with various water depths and bottom covers under which fish could hide. Exploration time was least in the tank with shallow water and bottom cover and highest in the tank with shallow water and no bottom cover. In choice tests and field observations, most fish preferred shallow areas with bottom cover. Thus, Sale concluded, there is no need to suggest the inheritance of complex cognitive maps and goal-directed behaviors, rather, the animal simply moves around more in an unsuitable habitat and less in a suitable one. Sale's model still does not explain how the animal "knows" what is suitable and what is not, or how stimuli from multiple cues are integrated. Nor does it explain the role of photoperiod (the duration of the animal's daily exposure to sunlight) in the response of dark-eyed juncos to photographs of their natural habitat. These wild-caught birds were presented a choice of viewing one of two 35-millimeter color slides showing different habitats. Birds kept in the lab under a winter photoperiod of nine hours of light and fifteen hours of darkness preferred (spent more time in front of) slides of their southern winter habitat. After day length was increased to fifteen hours of light and nine hours of darkness, the birds' viewing preferences shifted to the northern summer habitat. Social cues may also affect choice of habitat. Large juncos (usually males) dominate smaller individuals (usually females and juveniles) in wintering flocks. Biologist Ellen Ketterson explained the finding that females usually migrate farther south than males by hypothesizing that subordinate birds are forced to migrate farther to avoid competing with dominants. In their lab study, researchers E. Roberts and Peter Weigl found that during the short days (stimulating winter), small subordinate juncos showed the strongest preference for winter scenes. Risk of predation and competition are other factors that may affect habitat use. Hairy-footed gerbils live in vegetated islands in a sea of sand in the Namib Desert of southern Africa. Habitat use was determined by tracks in the sand and by how quickly they gave up feeding at stations containing seeds mixed with sand. Gerbils preferred sites around bushes or grass clumps to open areas and were more active on new-Moon nights than on full-Moon nights. They also gave up feeding at seed trays sooner in open areas and on full-Moon nights. These differences were likely caused by greater risk of predation in open areas and when the Moon was full. When striped mice, a close competitor of the gerbil, were removed, gerbils increased foraging activity, especially in the grass clumps. The immediate cues to which animals respond when selecting a habitat may not be the same as the ultimate factors that have brought about the evolution of the response. For example, the blue tit, a European bird, lives in oak woodlands where most of its preferred food is found. But the blue tit establishes its territory each year before leaves and caterpillars (its staple food) have even appeared, so it must be using some other cue, such as the shape of the trees, to select its habitat. In fact, we know little about the signals that animals respond to when choosing their habitat. And in migratory species, it is not even clear when in the life cycle a choice of habitat is made. One study found that breeding sites may be selected in late summer or fall before migration, rather than in the spring, as is usually assumed.