Background: Although the occurrence of malaria vector larvae in the valleys of western Kenya highlands is well documented, knowledge of larval habitats in the uphill sites is lacking. Given that most inhabitants of the highlands actually dwell in the uphill regions, it is important to develop understanding of mosquito breeding habitat stability in these sites in order to determine their potential for larval control. Methods: A total of 128 potential larval habitats were identified in hilltops and along the seasonal streams in the Sigalagala area of Kakamega district, western Kenya. Water availability in the habitats was followed up daily from August 3, 2006 to February 23, 2007. A habitat is defined as stable when it remains aquatic continuously for at least 12 d. Mosquito larvae were observed weekly. Frequencies of aquatic, stable and larvae positive habitats were compared between the hilltop and seasonal stream area using. chi(2)-test. Factors affecting the presence/absence of Anopheles gambiae larvae in the highlands were determined using multiple logistic regression analysis. Results: Topography significantly affected habitat availability and stability. The occurrence of aquatic habitats in the hilltop was more sporadic than in the stream area. The percentage of habitat occurrences that were classified as stable during the rainy season is 48.76% and 80.79% respectively for the hilltop and stream area. Corresponding frequencies of larvae positive habitats were 0% in the hilltop and 5.91% in the stream area. After the rainy season, only 23.42% of habitat occurrences were stable and 0.01% larvae positive habitats were found in the hilltops, whereas 89.75% of occurrences remained stable in the stream area resulting in a frequency of 12.21% larvae positive habitats. The logistic regression analysis confirmed the association between habitat stability and larval occurrence and indicated that habitat surface area was negatively affecting the occurrence of An. gambiae larvae. While An. gambiae and An. funestus larvae occurred throughout the study period along the streams, a total of only 15 An. gambiae larvae were counted in the hilltops, and no An. funestus were found. Moreover, no larvae managed to develop into adults in the hilltops, and the density of adult An. gambiae was consistently low, averaging at 0.06 females per house per survey. Conclusion: The occurrence of malaria vector larvae in the hilltop area was uncommon as a result of the low availability and high instability of habitats. To optimize the cost-effectiveness of malaria interventions in the western Kenya highlands, larval control should be focused primarily along the streams, as these are likely the only productive habitats at high altitude.