Scoring is where practice logs go wrong. Not the viewing, the scoring. Everything below is about making your own record mean something later, when you finally add it up.
A session produces a page of impressions. The target turns out to be a waterfall. Now you decide which of your impressions count. That decision is the entire experiment, and you're making it after you already know the answer.
Nobody is dishonest here. The mind fills gaps as a matter of course. "Movement" becomes a hit for falling water. "Cool" was obviously right. The three pages that pointed at a parking garage fade out of the account. Do that fifty times and you'll have a folder full of hits and no idea whether anything happened.
The fix is not trying harder to be fair. It's removing the moment where fairness is required.
Commit your impressions in a form you cannot edit, then look at the target. Not "then I wrote it up properly." Sealed, timestamped, done.
If you're on paper, write the session, then photograph the page before the reveal. The photograph is your record, not the page. This single habit removes most of the trouble, because retrofitting stops being possible rather than merely discouraged.
Free prose cannot be scored. "A sense of something vast and old and quietly powerful" matches a cathedral, a canyon, a redwood, a power station and a whale. It feels precise while committing to nothing.
Instead, reduce every impression to terms from a list you decided on before you started, and score only those. Here is the vocabulary Aperture uses. It is deliberately small.
The first, coarsest read. What kind of thing is it?
Then the detail, in eight categories. Roughly sixty words, and no others.
Synonyms are fine as long as they resolve to exactly one term and you fix the mapping in advance. "Gray" is "grey". "Freezing" is "cold". "Ochre" is "brown". What is never fine is a category-for-value swap: blue is not a hit for red, however close the feeling was.
Pick how many sessions you're running, and how many targets you're choosing between, before the first one. Write both down. Then run all of them, including the run where it's going badly and you'd rather stop.
This is the least enjoyable rule in practice work and the only one that makes the final number mean anything.
Deciding what counted once you've seen the target. Covered above, and it is by far the biggest one. Everything else on this list is a variation of it.
Checking your hit rate after every session and stopping the moment it crosses the line. Do this and you will cross the line eventually by luck alone. That isn't a signal, it's shopping for one.
The one where you were tired. The one where you got interrupted. The one that was obviously nothing. Every excluded session has a reason, and the reasons correlate with the result, which is exactly the problem.
"Energy" and "structure" will fit most targets if you let them. If a term would have counted for four of the five targets in the pool, it isn't evidence. Ask that question before you credit it.
Blue for red because both are colors. A bridge for a dam because both span water. Generosity here is the fastest way to a folder of hits that mean nothing.
Once the run is finished, the question is simple: is this hit rate doing something luck wouldn't? You cannot eyeball that. Six out of twelve on a one in four pool feels remarkable and isn't quite. Seven is.
Work out whether you're beating chance → Free, nothing stored, nothing sent.
Maybe it's noise. Maybe it isn't. This is how you'd actually find out.
Aperture is the practice journal this guide came out of: blind targets, a sealed impression, then the reveal, with the scoring done for you so it isn't yours to fudge. It's on the App Store. Leave an address if you'd like the occasional note when we have something worth reading.