1. The hypothesis was tested that intermediate connectivity to a river results in propagule inputs to wetlands, whereas excessive connectivity impedes recruitment, and insufficient connectivity causes less competitive species to be eliminated, with no recruitment of new species. As a consequence, very low or very high nutrient levels should decrease species richness by selecting specialized species, whereas intermediate nutrient levels should favour the co-occurrence of species with contrasting nutrient requirements. 2. Among cut-off channels with high sinuosity and which are infrequently flooded by the river (low flood scouring), one example possesses high species richness because most species are saved from extinction by long-term isolation of the channel and cold groundwater supplies. Other channels are poorly supplied with groundwater and show a lower richness of species, because of low propagule inputs and low recruitment potential. 3. Cut-off channels with low sinuosity and which are flooded at intermediate frequencies were divided into three groups. The first group was species-poor, being closely connected to the river through downstream backflows which maintain nutrient-rich and turbid waters, in keeping with the hypothesis. The second group presents intermediate richness caused by: (i) lower river backflows; and (ii) floods that partly scour substrate and plants, and afford regeneration niches for transported propagules. The third group was species-poor because of excessive groundwater supplies, which probably acted as a limiting factor for species growth and recruitment. 4. The most frequently flooded channel shows the highest species richness, and occurrence of rare and fugitive species, because of floods which compensate competition by scouring sediments and plants, and afford regeneration niches for propagules. In this case, conservation of biodiversity necessitates propagule sources at the level of the river landscape.