Neuromorphic computing has attracted enormous attention in recent years due to its impressive capabilities to address the inherent limitations of conventional integrated circuit technology, ranging from perception, pattern recognition to memory and decision-making. Inspired by the morphology and function of 1014 synapses in mammalian brain, silicon-based asynchronous spiking neural networks (SNNs) have ushered in an era of developing non-volatile resistive switching as memristors for artificial intelligence systems. Despite its low power consumption, high endurance and fast read/write time, two-terminal memristor suffers from the singular function and the lack of heterosynaptic plasticity, similar to the conditioned reflex of human beings. Inspired by advanced unconditioned reflex, multi-terminal memristive transistors (memtransistors) were developed to realize complex functions, such as multi-factors modulation and heterosynaptic plasticity. Here, we report the development of hybrid memtransistors, modulable by multiple physical inputs (light, field and electric bias), using hybrid perovskite and conjugated polymer heterojunction channels processed from solution-phase at room temperature. The devices exhibit non-volatile but highly reversible conductance modulation due to I-(Br-) ions migration which result to the formation of a tunable Schottky barrier at the injecting electrode interface. Meanwhile, the memtransistors show excellent gate tunability of six orders of magnitudes with large switching ratios (102), high endurance (>102 times) and long-term retention. Furthermore, photo-induced halides redistribution results in light plasticity of the polymer passivated perovskite memtransistor, which may be applicable in image identification and pattern recognition. Lastly, use of in situ scanning Kelvin probe microscopy and variable temperature measurement reveal the kinetics of the all-important bias- and photo-induced halide ions migration across the channel. Overall, the multi-physical input parameter memtransistor could enable the development of complex neuromorphic system capable of combining electrical with optical signals from the near-infrared to gamma rays.