Painting from memory is a different act than painting from reference. Not better or worse — different in kind. The hand moves differently. The eye does something different. The result carries a different kind of truth.
When you paint from reference, you're translating. The photograph or the still life or the model is the source, and your job is to move information from that source onto the canvas. You're solving a problem of correspondence. How close can I get? What do I need to simplify? Where does the translation break down?
When you paint from memory, you're not translating. You're reconstructing. And reconstruction reveals what you actually noticed — not what was there, but what registered. The things that didn't register simply don't appear. The painting becomes a map of your attention.
This is why memory paintings often feel more emotionally true than reference paintings, even when they're less accurate. They contain only what mattered. The irrelevant detail has been edited out by the act of forgetting. What remains is the essential thing — the quality of light that made you stop, the specific angle of a shoulder, the color that didn't match what you expected.
The relationship between memory and observation is something the technical side has been exploring from a different angle — the way that looking carefully at something changes what you remember about it. There's a feedback loop between attention and memory that painters know intuitively but rarely articulate.
Try this: paint something from reference, then put the reference away and paint the same subject from memory. The two paintings will be different in ways that surprise you. The memory version will be wrong in specific places — and those specific places are exactly where your attention failed. They're also often where the painting becomes interesting.
The reference version is more accurate. The memory version is more honest. Both are true. Neither is complete.
What you're really doing, in both cases, is making a painting about how you see. The reference just gives you more data to work with. The memory gives you less — and less, sometimes, is the more useful constraint.
More on the relationship between perception and mark-making at fine art practice. The question of what painters actually see versus what they think they see is also explored at tarot readings. And dev culture has been writing about the studio practice that makes this kind of experimentation possible.