Methods, systems, and devices for masked training and analysis with a memory array are described. A memory device may operate in a first mode in which a maximum transition avoidance (MTA) decoder for a memory array of the memory device is disabled. During the first mode, the memory device may couple an input node of the MTA decoder with a first output node of a first decoder, such as a first pulse amplitude modulation (PAM) decoder. The memory device may operate in a second mode in which the MTA decoder for the memory array is enabled. During the second mode, the memory device may couple the input node of the MTA decoder with a second output node of a second decoder, such as a second PAM decoder.