1. Introduced common wasps (Vespula vulgaris) reach high densities in the beech forests (Nothofagus spp.) of the northern South Island, New Zealand, and may be having a severe impact on populations of native invertebrates. An experimental approach was used to test whether reducing the abundance of common wasps increases the probability of native invertebrates surviving. Garden orb-web spiders (Eriophora pustulosa) were used because they were easy to collect and could be trained to build webs on a standard frame. Thirty orb-web spiders were placed out on web-frames in each of four study sites in beech forest during periods of low, medium and high wasp abundance, and their rate of removal was measured over a 4-h period. Wasp numbers were reduced by poisoning in two study sites. Using wasp-abundance data from weekly Malaise trap samples in the poisoned and non-poisoned sites, the probability of spider survival over the entire wasp season was modelled and the level of wasp control necessary to protect natural populations of orb-web spiders was estimated.

2. Wasp abundance and the probability of spider survival were negatively correlated, and smaller spiders were likely to survive longer than larger spiders. At the peak of the wasp season, significantly more spiders survived in the poisoned areas than in the non-poisoned areas.

3. The probability of a spider surviving to the end of the wasp season was virtually nil in the non-poisoned sites (5.08 × 10–18), but was also very unlikely in the poisoned areas (3.47 × 10–5).

4. The survival model predicts that wasp abundance would need to be reduced by 80–89.5% over the entire wasp season to protect populations of orb-web spiders.

5. Extrapolation from the model predicts that the invertebrate taxa most vulnerable to wasp predation may have already been removed from the beech forest ecosystem during the 40 years of wasp occupation.