It’s tough to do this when using the same images from prior systems. For a while you’ll probably find yourself actively thinking something like “was this the person or the object” and by pausing to consider this during memorization, it breaks your focus and flow. When your attention is taken off of the sequence of memorization and pulled to active consideration, it can be really tough to hold onto the previous information that you were trying to keep in short-term memory. I’ve found that as soon as I have to pause during an attempt and actively search my mind for the “correct” match, something in the prior scenes will get “bumped out” and it will really negatively affect the result.
So how is that avoided? Drill drill drill, exclusively with your “new” selected associations until the “alternate” image fades away. There’s really no way around it.
But the cool thing is, once you are fluent with your single association, I think its faster to encode than a PAO because you’ve eliminated a variable that you have to mentally consider during memorization. Even if you get comfortable with your PAO, there is still part of your recognition that has to parse whether a given number is supposed to be represented by a person or an action or an object. It’s like a flowchart that goes “what position is the number in (person or action or object) → if person, then this image → if action, then this image → if object, then this image.” But with a single association it just becomes “what is the number? → this is the image.” By cutting out those other decision paths, you ease the mental workload and can confidently and quickly pull the correct single associated image for each number. Then the only active work you need to do is to create the interaction and scene. The only downside to doing this with just a 2-digit system, is that repetitive imagery will become more common and that can cause potential confusion and swapping when the same image appears multiple times in an attempt. A 3-digit single association system will cut down drastically on the odds of repeats within a sequence, but again, that takes even more time and effort to build and learn.
One last note about number of loci vs. number of elements per scene that I think is worth emphasizing…
A lot of people get hung up on how many loci it takes to encode X number of digits and try all kinds of things to reduce the number of loci needed. This requires putting more elements into each scene. Some people expand the pao with other categories, like adjectives or quotes so that they can get 4x2 or even 5x2 digits per scene and use less loci. This is fine but in the end you still need to actively encode and decode the same number of total elements per attempt.
Compare that approach to simplifying your scenes and using more loci… If you only have to consider creating a scene with two intentional elements vs. three per scene, then yes you’ll need more loci. If you want to do 100 digits, you’ll need 25 loci at 4 digits per scene vs. 17 loci at 6 digits per scene, vs. 13 loci if you cram in 8 digits per scene… BUT, increasing the loci count should be fairly easy. Much easier than trying to build and recall much more complex scenes correctly, and those loci will be things that are effortless to remember. They’ll provide unique anchors for your now much shorter and simplified scenes.
I don’t know how many elements you’re trying to combine in each scene, but if you’re using 3 or more elements I’d encourage you to try simplifying it down to 2 and see how that feels. You may find that spreading your elements out across a few more loci gives you clearer and easier to build scenes than you get trying to make that extra connection between element 2 and 3 and so on within a scene. I know when I switched to only using two elements per loci, it made it SO much easier to recall my scenes. If you are confident with your memory palace sequence, the process becomes much less taxing. In my view the better option is almost always “use more loci” rather than “try to cram more stuff into fewer loci.”