The temperature and relative humidity inside the Glowworm Cave are spatially and temporally variable to a degree atypical of other caves in temperate latitudes. Comparison between the climatic data for 1977-80 and for 1955 shows that the climate of the cave was much more stable in 1955 than it is at present. This paper suggests that the increase in climatic variability is closely related to the fact that since 1975, when the Upper Entrance was unblocked, the cave has behaved as a wind tunnel: the difference in density between the atmosphere inside and outside the cave allows the outside air to flow freely through the interior between the Upper and Lower Entrances.Climatic fluctuations are considered to be the main factor responsible for the decline in the total population of Arachnacampa tuminasa (Skuse) observed between 1975 and 1980. Climatic changes may also explain why the numbers and diversity of other cave fauna are low.A. tuminasa has five larval instars. Larval development is homodynamic; the mqjority of the population follows a distinct annual cycle, in which most larvae hatch in spring. The distribution of glowworm larvae in the cave is largely determined by food supply and climatic conditions. The main source of food is the insects that emerge from the Waitomo Stream as it flows through the cave lakes. The main causes of mortality of glowworm larvae are cannibalism, an entomogenous fungus, cave opiliones, desiccation and flooding.

A freeze-coring device using liquid nitrogen is described, which enables one person to take a columnar core, extending from the surface to at least 50 cm below a stony streambed. An experiment to validate the technique showed that animals did not flee from the advance of the freezing-front. Using frozen streambed cores, the vertical distribution of benthic invertebrates of two streams in southern Ontario was investigated. In contrast to previous estimates,~70% of the fauna was found in the top 10 cm of the streambed, and invertebrate densities were often lower by an order of magnitude. These differences are attributed to problems of quantifying previous sampling methods.

No abstract