Multi-selection is a typical task where a set of targets is required to be acquired using serial (one-by-one) and parallel (group-by-group) approaches. The process of multi-selection on common 2- and 3-dimensional virtual interfaces is a typical task which has multiple well defined strategies. However, their suitability for use in an augmented reality context, which combines the real environment with virtual elements, requires further exploration. Due to the differences between the real and virtual environments, such as immovable parts of the environment and the lack of ”magical” movement techniques, a further examination of these commonly used selection strategies in this new context is necessary. The goal of this work is to perform an evaluation of the usability of certain lasso-based multiselection methods, which use a canvas to draw a selection shape. An evaluation of these techniques is performed in the context of an AR supermarket environment and will be used to outline their suitability for applications that specialize in analytical tasks. Based on research into related work, I designed and implemented several lasso-based multi-selection methods, which vary in form of use. I differentiate my implemented techniques by sorting them into distinct groups of ”embedded” and ”non-embedded” selection techniques. Embedded selection methods rely on a predefined static canvas for their selection, while the non-embedded implementations allow the user to dynamically reposition the drawing canvas for a selection. A constituent for both groups was evaluated in further detail by performing a small scale user study with 8 participants that focused on collecting mainly subjective data. The data collected during the user study suggests that the technique belonging to the embedded selection approach is a better fit than non-embedded selection approaches for this specific environment. Participants reported that embedded selection is more intuitive to use and required less of a learning phase to get comfortable with.