We use recent technological advances to simultaneously record brain signals from more than 100 recording sites in behaving non-human primates. From these brain signals, fast action potentials, also called spikes, can be extracted. However, each recording electrode captures the combined activity of many neurons as well as electrical noise from the environment, the recording system, and the brain. To better understand brain processes, it is important to capture the activity of individual neurons. For this reason, increasingly complex algorithms have been developed in previous years that make it possible to extract the activity of individual neurons from the combined signal of the recording electrodes. This process, called “spike sorting”, aims to identify the activity of as many neurons as possible with the highest possible precision. We have developed and use a spike sorting toolbox that employs novel nonlinear dimensionality reduction techniques to increase the yield and the accuracy of isolated single neurons from our recordings.