Non-native invasive plants can greatly alter community and ecosystem properties, but efforts to predict which invasive species have the greatest impacts on these properties have been generally unsuccessful. An hypothesis that has considerable promise for predicting the effects of invasive non-native plant species is the mass ratio hypothesis (i.e. that dominant species exert the strongest effects). We tested this hypothesis using data from a four year removal experiment in which the presence of two dominant shrub species (one native and the other not), and subordinate plant species, were manipulated in factorial combinations over four years in a primary successional floodplain system. We measured the effects of these manipulations on the plant community, soil nutrient status and soil biota in different trophic levels of the soil food web. Our experiment showed that after four years, low-biomass non-native plant species exerted disproportionate belowground effects relative to their contribution to total biomass in the plant community, most notably by increasing soil C, soil microbial biomass, altering soil microbial community structure and increasing the abundance of microbial-feeding and predatory nematodes. Low-biomass, non-native plant species had distinct life history strategies and foliar traits (higher foliar N concentrations and higher leaf area per unit mass) compared with the two dominant shrub species (97% of total plant mass). Our results have several implications for understanding species’ effects in communities and on soil properties. First, high-biomass species do not necessarily exert the largest impacts on community or soil properties. Second, low-biomass, inconspicuous non-native species can influence community composition and have important trophic consequences belowground through effects on soil nutrient status or resource availability to soil biota. Our finding that low-biomass non-native species influence belowground community structure and soil properties more profoundly than dominant species demonstrates that the mass ratio hypothesis does not accurately predict the relative effects of different coexisting species on community- and ecosystem-level properties.