Behavioral Ecology Group, Department of Zoology, The Ohio State University, 1735 Nell Avenue, Columbus, Ohio 43210, USA Risk of starvation and risk of predation are two major evolutionary forces thought to shape an ani- mal's foraging behavior (Lima 1986, McNamara and Houston 1990). If energetic demands are high, the risk of starvation could be high, and individuals should increase their energy reserves by foraging more intensively. If the risk of predation is high, in- dividuals should become more vigilant to detect predators. However, when vigilant for predators, in- dividuals could be forced to reduce other activities such as foraging (e.g. Caraco 1979, Elgar 1989). McNamara and Houston (1986) modeled vigilance behavior in diurnal animals as a function of energy reserves and time of day. They predicted that vigi- lance rate should be low in the morning, a time when animals replenish fat stores lost during the nocturnal fasting period. Vigilance also was predicted to be low late in the day as animals accumulated energy reserves for the following night. In contrast, the re- duced demand for energy gain around midday was predicted to allow a heightened vigilance rate. We are aware of no empirical tests of these predictions. In the present study, we examined the relationships among vigilance rate, energy reserves (body mass), and time of day in Tufted Titmice (Baeolophus bicolor). We also examined the effects of ambient temperature because vigilance has been shown to be positively  Present address: Department of Biological Sci- ences, Purdue University, West Lafayette, Indiana 47907, USA. E-mail: vladimir@bilbo.bio.purdue.edu correlated with temperature (Pravosudov and Grubb 1995). Methods.--We studied vigilance in six male Tufted Titmice during the winter of 1994-95. The birds were caught at different locations in central Ohio and ob- served in isolation for eight days in an outdoor avi- ary 3 x 8 x 2 m high. During that period, food (un- shelled sunflower seeds) and water were provided ad libitum. The aviary was open to the sky (except for wire netting) but was walled by translucent fiber- glass panels that isolated the birds visually and maintained wind speeds within the aviary at or near zero. One end of the aviary was roofed by a trans- lucent fiberglass panel 30 cm wide that protected a "recording" perch and feeder from rain and snow. We observed the titmice from an observation chamber attached to one end of the aviary. Obser- vations occurred under prevailing photoperiods, and ambient temperatures were recorded and stored ev- ery 30 rain in a computer housed in an attached ob- servation chamber (Weather Wizard III and Weath- erlink software; Davis Instruments). An electronic balance inside the observation chamber was con- nected through a one-way glass wall to a perch used by the birds frequently during the day and for roost- ing at night. Readings from the balance were record- ed to the nearest 0.01 g and stored in another com- puter, also housed in the observation chamber. To evaluate a bird's vigilance, we divided each day into four 2-h blocks from 0900 to 1700 EST. During each block, we randomly selected one 30-rain period during which we observed vigilance. We obtained a measure of vigilance by recording the number of times a bird "looked up" while perched on a hori- zontal branch handling and eating a sunflower seed held between its feet. We considered a bird to be looking up when its bill was pointed above the hor- izontal (Pravosudov and Grubb 1995). Although birds sometimes may have looked up while swallow- ing food pecked from a seed, they also looked up pe- riodically while removing the shell to obtain access to the seed. Thus, we assumed that looking up dur- ing opening and eating of a sunflower seed repre- sented interruption in food intake related to being vigilant. For parids such as the Tufted Titmouse, measuring vigilance while the bird is handling a food item seems to be more valid than while a bird is foraging, during which it also could be looking up to scan for prey items or for another foraging site. For every observation period, we calculated a mean rate of vigilance (look-ups per s) based on 4 to 30 observations. Thus, we obtained four measures of vigilance per day (one in each time block) for eight days for each bird. Also for each time block, we ob- tained mean body mass and average ambient tem- perature. Statistical analysis.--To test for the effects of body mass, ambient temperature, and time of day on vig- ilance, we used a repeated-measures ANCOVA with time of day (four levels) as a factor and body mass and temperature as covariates. To investigate wheth- er there was a directional change in vigilance during the day, we also used a nonparametric repeated- measures, ordered alternatives Friedman test for the effect of time of day on vigilance (Hollander and Wolfe 1973). For this analysis, all vigilance records for each bird were reduced to one per time block per day. Results.--There were significant, positive relation- ships between titmouse vigilance and both ambient temperature (b = 0.002, t = 2.17, P = 0.03) and body mass (b = 0.07, t = 5.61, P < 0.001). After controlling for these covariates, we also found a significant as- sociation between vigilance rate and time of day (F = 3.23, df = 3 and 15, P = 0.05), with titmice steadily increasing their vigilance throughout the day (Fried- man test, L = 179, n = 6, k = 4, P < 0.001; Fig. 1). None of the first-order interactions among the two covariates and the factor was statistically significant. Another repeated-measures ANCOVA, with the bird as a random factor and ambient temperature, body mass, and time of day as covariates, produced simi- lar results. Each of the three covariates was positive- ly and significantly related to vigilance (multiple R 2 = 0.67, F = 47.02, df = 3 and 176, P < 0.001; tem- perature, b = 0.002, t = 2.09, P = 0.03; body mass, b = 0.07, t = 5.52, P < 0.001; time of day, b = 0.02, t = 3.60, P < 0.001). None of the first-order interactions was statistically significant. All of these analyses suggested that the three variables affected vigilance rates independently in Tufted Titmice. 1.00 ' 0.96 , 0.92 o - 0.88  0.84  0.80  0.76  0.72  , ,  0900-1100 1300-1500 TIME OF DAY Fc. 1. Relationship between vigilance rate and time of day in six Tufted Titmice. Values are mean _+ 1 SE. Discussion.--That titmice were more vigilant in warmer weather is not surprising. A positive rela- tionship between vigilance rate and ambient tem- perature has been demonstrated previously for free- ranging Tufted Titmice (Pravosudov and Grubb 1995), Willow Tits (Parus montanus; Hogstad 1988), and Yellow-eyed Juncos (Junco phaeonotus; Caraco 1979). In lower ambient temperatures, a bird must in- crease its metabolic rate to maintain body tempera- ture. The higher metabolic rate demands more en- ergy, which requires more intensive foraging. In or- der to increase its foraging activity at lower ambient temperatures, a bird may compensate by decreasing its vigilance rate. Caraco (1979) and Hogstad (1988) demonstrated the effect of temperature on vigilance rate indirectly through a change in group size which, in turn, was related to ambient temperature. By con- trast, Pravosudov and Grubb (1995, this study) found evidence for a direct influence of ambient tem- perature on vigilance rate, independent of group size. There was a positive relationship between body mass and vigilance rate within the birds. This find- ing supports the prediction of McNamara and Hous- ton (1986). Contrary to McNamara and Houston (1986), however, our titmice increased their vigilance rate throughout the day independent of ambient temperature and body mass. We can think of two possible explanations for this result. The first is based on the assumption that a bird with consider- able energy reserves, regardless of possible mass-de- pendent predation risk, can afford more time to be vigilant for predators. The second follows from the assumption that a heavier bird is more easily caught by a predator. Vigilance rate was positively correlated with body mass independent of time of day. Therefore, when a bird had relatively large fat reserves in the morning, it was more vigilant than when morning reserves were relatively low. We also found that all of the tit- mice increased their body mass (fat reserves) steadi- ly throughout the day, so that the birds always were lightest in the morning and heaviest in the evening (Pravosudov and Grubb 1997). Therefore, it is pos- sible that vigilance rates were in response to a bird's average level of energy reserves rather than to time of day. McNamara and Houston (1986) argued that animals should be less vigilant in late afternoon be- cause, as evening approaches, time devoted to ac- cumulating energy reserves for the night becomes of paramount importance. However, a safe level of en- ergy reserves for the night already may be achieved sometime nearer midday, particularly in an experi- ment like ours where birds have access to food ad li- bitum. The other explanation for the positive relationship between vigilance and time of day derives from the assumption that heavier birds are more vulnerable to predation (e.g. Lima 1986, McNamara and Houston 1990). It has been assumed that heavier birds have higher predation risk because they are less maneu- verable and fly more slowly than lighter birds (Wit- ter et al. 1994, Metcalfe and Ure 1995). A heavier bird under higher predation risk may employ a higher vigilance rate to reduce such risk. Because our tit- mice always were lightest in the morning and heavi- est in the evening, it is possible that during the late afternoon heavier birds were more vigilant to com- pensate for their higher perceived predation risk at that time. We have described these two explanations for increased vigilance as if they were independent. Of course, it also is possible that they operate to- gether in causing vigilance to track body mass throughout the day. In conclusion, because a high level of vigilance re- duces the risk of predation but also lowers the rate of energy gain, knowledge of how animals set their vigilance levels is important for an understanding of daily energy routines. In mathematical models of en- ergy management, vigilance should be included as a constraint on energy gain. Acknowledgments.--We thank C. L. Bronson, Paul Doherty, Bob Mauck, and Elena Pravosudova for technical assistance, and three anonymous reviewers for comments on the manuscript. This project was partially funded by an Ohio State University Grad- uate Alumni Research Award and an O.S.U. Presi- dential Fellowship to V.V.P., and by National Science Foundation Grant IBN-9522064 to T.C.G. The work was carried out under Ohio State University animal use protocol 94A144. LITERATURE CITED CARACO, t. 1979. Time budgeting and group size: A test of theory. Ecology 60:618-627. ELGAR, M. A. 1989. Predator vigilance and group size in mammals and birds: A critical review of the empirical evidence. Biological Reviews of the Cambridge Philosophical Society 64:13-33. HOGSTAD, O. 1988. Advantages of social foraging of Willow Tits Parus montanus. Ibis 130:275-283. HOLLANDER, M., AND D. A. WOLFE. 1973. Nonpara- metric statistical methods. John Wiley and Sons, New York. LIMA, $. L. 1986. Predation risk and unpredictable feeding conditions: Determinants of body mass in birds. Ecology 67:377-385. MCNAMARA, J. M., AND A. I. HOUSTON. 1986. The common currency for behavioral decisions. American Naturalist 127:358-378. McNAMARA, J. M., AND A. I. HOUSTON. 1990. The value of fat reserves and the trade-off between starvation and predation. Acta Biotheoretica 38: 37-61. METCALFE, N. B., AND S. E. URE. 1995. Diurnal vari- ation in flight performance and hence potential predation risk in small birds. Proceedings of the Royal Society of London Series B 261:395-400. PRAVOSUDOV, V. V., AND T. C. GRUBB, JR. 1995. Vig- ilance in the Tufted Titmouse varies indepen- dently with air temperature and conspecific group size. Condor 97:1064-1067. PRAVOSUDOV, V. V., AND GRUBB, T. C., JR. 1997. Man- agement of fat reserves and food caches in Tuft- ed Titmice (Parus bicolor) in relation to unpre- dictable food supply. Behavioral Ecology 8:332- 339. WITTER, M. C., I. C. CUTHILL, AND R. H. BONSER. 1994. Experimental investigations of mass-de- pendent predation risk in the European Starling, Sturnus vulgaris. Animal Behaviour 48:201-222. Received 7 February 1997, accepted 4 June 1997. Associate Editor: K. L. Bildstein