There is a lot of research on 'Cognitive Overload' (see Wickens’ (2002) Multiple Resource Model). Verbal analysis appears to be a 'single channel' in the human brain. This means that you can read and understand or listen and understand but you cannot do both simultaneously. So someone who is concentrating on reading say the check list - or even visualizing the check list and verbally analyzing the memory - will appear not to hear what is said to them, even STALL STALL STALL. There is a similar issue with spatial analysis so if you are listening to someone on your cell phone while driving and they start giving you directions your brain will start processing the spatial analysis from the phone call - and will stop processing the visual cues from the road you are driving along; this is one of the reasons talking on the phone is dangerous.
More interesting still these cognitive channels also have a level of interference with each other. So if you are attempting to regain straight and level in Alternate law watching the various instruments and trying to listen to the PNF reading checks you will miss a lot of what is being said.
I do not believe that sufficient thought is given to cognitive overloads in the design of instruments and flight deck procedures: add the startle effect and 'automation surprise' plus the sudden stress of an aircraft that appears to be behaving differently to the instruments - and it will appear that the PF is ignoring everyone and auditory cues. This is a human limitation. If you don't believe it recite a nursery rhyme and as you do re-read this posting and have someone talk to you and try to understand what they have said. You cannot do it. The only input that will break into this lock out is 'haptic' that is touch - smack the guy on the arm and that will usually break attention from the other cognitive inputs