One of the most thought provoking experiences that I have ever witnessed was in a Boeing 737-200 simulator where I was conducting type rating training on a highly experienced SE Asia born airline captain.
He was asked to intercept a VOR radial while climbing at 250 knots on a departure. He was on autopilot. The first officer had only 300 hours total time. Both pilots displayed some lack of familiarity with RMI interpretation.
Asked to now hand fly and level out at 8000 ft, the captain showed lack of familiarity with basic cross-reference skills. Without warning his ADI failed (became frozen) during a level turn but neither he or his F/O noticed this until the captain was asked to take up a new heading.
Despite a comparitor light showing , the captain steadily wound on ever increasing angle of bank (ADI was "frozen) while attempting to turn to the new heading. The standby ADI and the F/O ADI were operating normally. The simulator was set on night IMC scene.
With the "frozen" ADI still not responding to his roll input, the aircraft became inverted due to the captain's continued roll inputs.
The nose dropped and at this stage the captain looked puzzled as he felt that obviously something was not quite right.
The F/O's eyes were staring at his own ADI and he seemed frozen in fear as the IVSI went off the clock. He sat on his hands and said nothing because he simply was out of his depth.
Passing through around 120 degrees angle of bank and 20 degrees nose down, the captain suddenly looked at the standby ADI (correct) indication, and disbelievingly called out "Standby ADI failure!" He then proceeded to reach over and pull the standby ADI caging knob. The no doubt astonished standby ADI did what it was told and showed "level" flight under the influence of the pulled caging knob. After several thousand feet of altitude had been carved off it was clear that the plot was lost, so the instructor froze the simulator in order to knock off for morning tea and a little chat.
Automatics complacency can do those things to some pilots.