Abstract: In one embodiment, a method for detecting autism in a natural language environment using a microphone, sound recorder, and a computer programmed with software for the specialized purpose of processing recordings captured by the microphone and sound recorder combination, the computer programmed to execute the method, includes segmenting an audio signal captured by the microphone and sound recorder combination using the computer programmed for the specialized purpose into a plurality recording segments. The method further includes determining which of the plurality of recording segments correspond to a key child. The method further includes determining which of the plurality of recording segments that correspond to the key child are classified as key child recordings.