The presence of Homo floresiensis and Homo luzonensis in southeast Asia 90,000 to 60,000 years ago is considered surprising by many, and has been used to support their designation as unique species and the islands they were discovered on as refugia. Here, we statistically test the null hypothesis that H. floresiensis and H. luzonensis represent temporally uninterrupted occurrences relative to Homo erectus. We do this using the 'surprise test' for the exceptionality of a new record. Results demonstrate that H. floresiensis and H. luzonensis are not temporally distinct relative to H. erectus. Their late persistence should, therefore, not be considered surprising, they cannot reliably be inferred to be outside of H. erectus' temporal range, andtemporallythe islands of Luzon and Flores are not supported as refugia. Similarly, late H. erectus at Ngandong, Java, is not demonstrated to be temporally distinct relative to earlier, principally mainland-Asian, H. erectus. Further, we demonstrate that substantial numbers of fossil discoveries would be needed before H. floresiensis and H. luzonensis are outside of H. erectus' expected temporal range. If H. floresiensis and H. luzonensis are descended from H. erectus populations, our results point toward either geographic processes of allopatric speciation or behavioural processes leading to a sympatric speciation event.