Within the field of numerical cognition, concept of cognitive magnitude refers to a biologically determined ability to comprehend, approximate and manipulate different quantities. This paper set a goal to determine the effects, scope and differences between different mechanisms underlying numerical and spacial estimations of magnitudes. Experimental procedure was used to test absolute error, time of response, and meta cognitive estimations of response confidence in a total of 56 different experimental situations. These situations where formed based on three independent variables: (1) group (group with a visible referential object, and a group without a visible referential object), (2) task (spacial assessment of magnitude, and numerical assessment of magnitude), (3) difficulty level (1 – 14). All participants where randomly placed in one of two groups, in which they solved 28 task examples (14 examples per task type). Task of spacial assessment of magnitude required participants to change the size of a referential frame to fit all targeted objects, while the task of numerical assessment of magnitude required participants to estimate how many target objects can fit within the referential frame. Results show that spacial assessment tasks result with smaller absolute errors, underestimation of magnitudes, longer reaction time and lower meta cognitive estimations of response confidence in comparison with numerical assessment tasks, where the trends for examined variables were opposite — greater absolute errors, overestimation of magnitudes, shorter reaction time and higher meta cognitive estimations of response confidence. These features became more pronounced for both tasks as difficulty increased. Differences between groups where found only for one variable - reaction time. It was found that the group without a visible referential object was faster in assessing magnitudes than the group with a visible referential object.