It’s worth noting up front that we find our current methods of implementing memory palaces to be relatively simple (see “Doesn’t it take too much time to make all these palaces?”). We don’t think that using palaces in a mnemonic strategy adds up to much extra time or effort. Otherwise, the following might be moot. The presence of this question is also not to say that standalone images are never useful. We often create them for out-of-context facts or when it may be useful to create quick images on the fly. When it comes to carefully learning structured material, however (e.g. I'm sitting at my computer to learn the lung pathology chapter), I’ve found there to be three main arguments in favor of palaces:
1. Organization: Palaces can be a boon to the structuring of stored information. I try to use the inherent structure of palaces to keep particular subtopics neatly contained within a given space in the mind. Let’s say I’m using my high school swimming pool as my palace for the lung pathology chapter. I reach a section about interstitial pneumonia, a key type of pneumonia. I might choose the men’s locker room as my exclusive “area” for interstitial pneumonia (this room was simply next along my path through the pool as I progressed through the subtopics of lung pathology). I can tidily contain everything worth memorizing about interstitial pneumonia within this space. When then thinking about interstitial pneumonia, I can easily “teleport” to this one room, which I know contains everything I want to remember. Although often rooms, an “area” might be an entire building (e.g. a neighbor’s house, the details of which I don’t know well), a field in a sports complex, a section of a park, a street, a parking lot, etc. Check out our Tetracyclines and Acute Pyelonephritis videos for examples of this in action.
I find this organization incredibly helpful. Here’s how I’ve noticed my thinking tends to go, and I believe it’s a useful practice: On hearing a topic, I first consider what semantically springs to mind and what intuitive implications arise based on the mechanisms at play (e.g. “Ok, interstitial pneumonia, being spread out and ‘atypical,’ displays weaker symptoms—lower-grade fever and less sputum—than do other pneumonias”; finding this intuitive and easy to remember, I didn’t encode it). As my natural memory begins to run out, I scan the relevant area to cue me into the different things I originally chose to remember (e.g. the specific bacteria and viruses which cause it, which aren’t exactly intuitive). In this way, I force myself to think conceptually while using the palace to cue me into critical info I might not have readily recalled (e.g. “Oh yeah, because interstitial pneumonia weakens immune defenses, it can often present with superimposed Staph or H. flu bronchopneumonias, which are deadlier”). This organizational approach also provides a framework for adding new info I might learn down the line (e.g. I later learned that Coxiella is an additional bacterial cause of interstitial pneumonia, so I added an image to a new locus within my original locker room area). The bottom line: palaces keep your images for each subtopic nice and organized. Unless you create incredibly complex images (which are less effective), you don’t get that with free-standing images.
2. Sequence: Using palaces is an easy way to track the progression of knowledge. By encoding information along a path, you can get a deeper understanding of how each topic builds upon its predecessors. It also allows you to easily step back and survey a broader topic (e.g. “Ok, when it comes to lung pathology, I’ve learned nasopharyngeal issues, laryngeal issues, pneumonias, effects of TB, obstructive and restrictive diseases, neoplasms, etc.”). Again, you don’t get that with free-standing images.
3. An extra hook (i.e. a unique place in the mind): By encoding info at a unique locus, you give each piece an extra association—a special position in the mind. As we’ve discussed elsewhere, humans have an instinctive knack for visualizing physical spaces, so associations with locations are incredibly robust. Let’s saying I’m trying to learn the Spanish word for “to rent,” which is “alquilar.” I might simply imagine pouring some alkaline milk onto a taxi (the first things I thought of relating to “alquilar” and “rent”). Why not stop there, with these images free-floating in the ether? You could, of course, and this would still be helpful (this is how Memrise works, roughly speaking). But I think we can do better. By placing this milk-taxi on top of a table in my college dining hall (the position I was at in my palace when learning this word), I give it an extra hook. When I hear the word alquilar in the future, I find myself involuntarily teleporting to this locus, often without even bringing the milk-taxi visual to mind. I often find the location matters more than the images themselves (see “How clear should my visualizations be?” for additional discussion), and I’ve heard many echo this sentiment. Alternatively, you might simply affix this milk-taxi to a specific mental place (e.g. a particular street corner), without it being along any particular mental path or palace. This similarly confers a spatial association and therefore works well (this is how I do the historic dates event at memory competitions). For learning applications, however, I’ve tended to follow the former method of using paths through palaces, mainly for the reasons discussed in #1 and #2.
To sum up, I’ve found that memory palaces can be implemented painlessly while conferring a host of benefits. I didn’t always feel this way, initially finding memory palaces to be clunky long-term learning tools. Since implementing our highest-yield adjustments, however (see “Do memory palaces hinder learning?”), my belief in the above ideas has continued to grow.