Yes, all of these!
Its important to practice independently associating the cards with their People, and then their Actions, and then their Objects. You should strive towards instantaneous recognition of each, without having any extra conversion steps.
More on this here (note that it references PAO for numbers, but it equally applies to cards):
Try not doing this. Its really really tempting, but if you can not do re-runs it will help you tremendously. You’ll see your accuracy take a hit initially, bit you’ll be learning to trust your first look visualizations and you may learn what cards give you consistent trouble so you can work on reviewing those.
I’d be willing to bet this means that you flip over 3 cards at once and read them one by one as a scene, right? If thats the case, the single card metronome approach would still work fine because on every click you need to move on to reading the next element. Person “click” does Action “click” with/to Object “click”, visualize it “click” next set…
As a sidenote, this thread gets sort of into semantics about elements, images, and scenes but you might find it interesting and helpful when it comes to thinking about data compression and the efficiency of a system: Reason for deleting 10 digit system post - #6 by TheHumanTim