This paper investigates the improvement of learning sensorimotor models for developmental robots, in particular robot arm kinematics models, with inter-robot knowledge transfer. Developmental robots progressively learn through embodied interaction with the physical environment. In the single-robot case, exploration in the world is performed in isolation and the robot explores its own capabilities. In a multi-robot case, with one or more experienced robots, we argue that it may be beneficial for the robots to be able to share the knowledge they have acquired through their individual exploration. We explore knowledge transfer in the context of learning arm kinematics models, where an experienced robot shares its kinematic data with a new robot that is autonomously exploring its environment. We show that the sensorimotor models of the new robot can be bootstrapped by the shared knowledge, converge faster and also achieve a better asymptotic performance compared to individual exploration from scratch. We perform an analysis of knowledge transfer in simulation, ranging from simple two-link planar robots to redundant systems.