Abstract Native trees from the Caribbean were tested for seed desiccation responses, by adapting the “100-seed test” protocol. Ninety-seven seed lots of 91 species were collected in the Dominican Republic and tested for germination immediately after collecting, and after drying and moist storage. Seed desiccation sensitivity was assessed as a continuous variable (Viability Loss Index; VLI), based on seed germination values before and after drying. The results were compared with predictions of seed desiccation responses based on seed lot traits (initial moisture content and thousand-seed weight) and with those of published predictive models based on plant and seed traits. VLI could be calculated for seed lots of 40 species. 80% of these seed lots showed consistent results among experiments and predictive models. Issues on the set up of the experiments were discussed, as well as the species for which experimental results and predictions led to contrasting results. Overall, the “100-seed test” confirmed to be an effective tool for assessing seed desiccation responses of a diverse under-investigated woody flora, guiding the seed conservation of trees and their use in reforestation programmes. In addition, by providing new data, it might improve the performance of available predictive models.