Recently, researchers have been trying to determine whether it may be possible to differentiate two types of dementia, namely the behavioral variant of frontotemporal dementia (bvFTD) and Alzheimer's disease (AD), based on the ability to use theory of mind, and have used various theory of mind tasks for that purpose. The main aim of this study was to examine the performance of people with bvFTD, people with AD, and a control group consisting of elderly people without dementia on a task using a continuous measure of theory of mind called the Sandbox Task. This study examined Le Bouc and colleagues’ (2012) assumptions that people with bvFTD have a specific theory of mind deficit in inhibiting their own perspective, whereas people with AD have a deficit in reasoning about others’s beliefs. A further assumption was examined in the control group, namely that the manipulation of removing the target object may affect the participants' performance in the task. The study included 6 people with bvFTD, 9 people with AD and 64 elderly people without dementia. In the Sandbox Task, participants observe the interaction between person A and person B, where person A first hides the object in one location and then, when person A either leaves (false belief conditions) or remains present (true belief condition), person B moves the object to another location. In the test trials, participants are asked where will person A look for the object, and in the memory control trials where did person A hide the object. The Sandbox Task was presented to participants as a paper booklet depicting different scenes and they manually marked their markings for the location of the object. Participants' performance was recorded based on their egocentric bias indices, which represent the difference in the difference between the location marked by the participant and the exact location of the object between test trials and memory control trials. In the control group, the assumption that removing the target object from the scene reduces participants’ egocentric bias was not confirmed. As for participants with bvFTD and AD, the assumptions of Le Bouc and colleagues (2012) were for the most part not confirmed. However, due to the small sample size in these two clinical groups, it is not possible to draw strong conclusions. Therefore, larger clinical samples should be included in future studies. Likewise, further research is needed to determine whether the Sandbox Task yields an adequate measure to test theory of mind in clinical groups and whether this task can provide some useful explanations about the mechanisms underlying the potential theory of mind deficits in people with these two types of dementia.