Historic deforestation has deprived many river systems of their natural wood loadings. To study the effects of the loss of wood from waterways, a field trial was conducted in three small forested streams in New Zealand. The objectives were to (i) examine differences in fish assemblages among wooded pools (where wood provided cover), open pools and riffles and (ii) measure the effects of wood removal on channel morphology and fish assemblages. In the first part of the study, no significant differences were found in total fish density among the three habitats. However, total fish biomass was significantly higher in wooded pools (64% of total fish biomass) compared with open pools and riffles. Mean density and biomass of banded kokopu (Galaxias fasciatus) and mean biomass of longfin eel (Anguilla dieffenbachii) were highest in wooded pools, whereas the density and biomass of bluegill bully (Gobiomorphus hubbsi) and torrentfish (Cheimarrichthys fosteri) were highest in riffles. In the second part of the study, wood was removed from a 200-m section (treatment) in each stream, significantly reducing pool area and increasing the proportion of channel area and length in riffles. At the habitat scale, banded kokopu and large longfin eel were the two species mostly affected by wood removal. At the reach scale, banded kokopu biomass was significantly lower in the treatment sections. Although wooded pools were a small portion of total habitat, they provided important habitat for two of New Zealand's larger native fish taxa.