The eastern Kendeng Hills (Java, Indonesia) expose a 1000 m thick series that is used as a stratigraphic standard, representing the emergence of eastern Java from the sea. The fluvial top is rich in vertebrate fossils and yielded the Mojokerto (Perning) hominin skullcap, which is regarded as the earliest evidence of Homo erectus on Java, with age estimates ranging between 1.9 and 1.49 Ma. The series is commonly regarded as an uninterrupted record of coastal progradation. However, recent studies show that the emergence of eastern Java has been a complex process, under influence of tectonism, volcanism, sea-level fluctuations and fluvial dynamics, leaving a fragmented depositional record that varies from site to site. This is at odds with the prevailing stratigraphic practice of long-distance correlations and questions the existing interpretations of the eastern Kendeng reference sections. Here we present the results of a fieldwork-based re-interpretation of this key stratigraphic record, which we identified as the fill of a previously unrecognized Plio-Pleistocene embayment, surrounded by elongate uplift zones. Clinoform-bedded sandstones relate to a stage of explosive, high-silica volcanism, supplying large volumes of ash. The embayment fill is incised and covered by fluvial deposits, which we relate to the Middle Pleistocene Brantas. The fluvial strata have a cyclic build-up, probably representing sea-level controlled stages of aggradation and degradation. Based on a reconstruction of fluvial cycles, we provisionally link the conglomerate bed in which the Mojokerto Homo erectus was found to MIS14 (∼550ka). We infer that the published radiometric ages derive from reworked volcanic clasts that make up this incisive fluvial lag and are not representative for the age of deposition. Our study places the eastern Kendeng series in a new landscape context and changes our view of the timing of hominin migration to Java.