Humanoid social robots have been researched and deployed in fields such as healthcare, shopping, and telecommunication. These are all meant to mimic human behaviour, speech and interaction type to some degree in order to better fit in the role they are taking on. Other successful examples of social robots have been constructed by copying animal behaviour and selected social cues instead.The levels of comfort, trust, positive perception, and reliability a robot emits to a person is correlated with its appearance and actions. Depending on the human's experience, past, and knowledge, as well as biological classification, the levels are furthermore impacted independently of the robot.This has lead to many taken-for-granted studies on human-like HRI, which are hard to replicate at a high fidelity and findings are not always reproducible. Moreover, non-humanoid robots have more options for innovation and robot-specific communication and interaction features design, as they are freed of the constraint requiring them to copy existing living beings. Additionally, considering non-verbal communication, they are forced to rely on social cues depending on motor skills, poses and gaze or light and sound signaling to make themselves perceivable, legible, predictable and socially acceptable. Otherwise, having this much freedom of interpretation, replicating studies involving non-humanoid robots in non-verbal constrained scenarios proves to be very delicate. The motivation and scope of this master thesis is therefore to address, evaluate, and expand upon the two main topics presented so far, namely the non-verbal communication and interaction between humans and non-humanoid robots while considering the limitations of past studies and the phenomena of the replication crisis in this field of study.The main work of this research consists of replicating a previous laboratory study and expanding upon its design and findings. In the original experiment, young adults were asked to interact with non-humanoid robots, called Sphero, which, having only lights, sounds and rotatory movements at their disposal, had to make themselves understood to the participants when asking them to follow it to a reward. Furthermore, this scenario is expanded upon by including also older adults in the experiment and comparing the results to reduce bias. Moreover, to explore a broader range of human-robot interaction (HRI) with a non-verbal non-humanoid robot, a second scenario was also integrated, where the reactions of participants to the robot suddenly bursting out in a dance to music are sought after.The key findings of this thesis are of a dual nature.On the one hand, it shows that obtaining the close to the same results with both the young and the old participants individually and combined, during the first scenario run-through of the original study, was impossible, considering the limitations of the materials, environment and personas.Thus, the replication crisis is supported as an existing and occurring phenomena in HRI.On the other hand, the results from the added second scenario incorporating dancing further shed light on the attitude of participants towards the robot as a companion versus a care-taker, prove the effect of novelty and its influence on the different age groups, as well as the emotional response and implication, varying from one generation to another, beyond differences between individuals themselves. Moreover, established and innovative ways of non-verbal communication attributed to a non-humanoid robot were researched, tested and offer a basis for future research considering non-humanoid robots at large.This thesis is relevant as it investigated and innovated the area of non-verbal communication, considering non-humanoid robots with reduced and constricted mechanical and functional capabilities. Furthermore, it took one step into the direction of assessing, processing and solving the replication crisis in HRI, by exploring its occurrence in a consecrated study example. Lastly, but not least, it helped to better understand the human interlocutor in interaction contexts involving non-humanoid social robots, validating learnings on the design of robots and their behaviour, as well as on the common misconceptions and staple statements regarding individuals classified by age, gender, and experience.