In the learning set, you cannot exit without completing the entire activity or shutting down the app. Also, my kids would just match the visual cues in the game, not tune into the words and their meanings. For example "high" and "low both have a truck, so they would pick those because they are both trucks, not because they know the opposite meanings of high/low. Same with empty/full - both bottles. All the pictures I saw in this sample were like this. I have my doubts as to the learning value in this. I would challenge you to turn off the sound on the app and I can guarantee they will still match the pictures based on the objects, not the attributes. Possibly give the stimulus item verbally only, without the picture to match to?