studywren.

Dictionary discipline: the three-pass rule

2026-08-06 · languages

My vocabulary list used to be two hundred entries long, of which I actively used maybe fifteen. The list was not the problem. The problem was that every unknown word got written down the moment I met it, which means the list was really a record of things I had seen once, and seeing a word once is not information about whether you will ever need it.

The fix has been one rule for a year: a word must be looked up in three separate sessions before it is allowed into the list. Not three look-ups in one sitting — three encounters, days apart, in different texts. If it comes back a third time, the language clearly wants to teach it to me, and the entry gets made in the minute it takes to write a sentence with it.

What the attrition looks like

Roughly two-thirds of looked-up words never survive to pass two. That is not waste; that is the filter working. Those words were noise — a register I will never use, a topic area that is not mine, a one-off in one article. Cutting them out is the entire efficiency gain: my list is now under sixty entries and I know essentially all of it, instead of being technically familiar with two hundred things and able to deploy almost none.

The rules the rule needs

Look-ups still happen freely at pass one — hesitation in the middle of a text is expensive, so the dictionary stays instant and only the notebook is rationed. Names and technical terms in the reading are exempt from the whole scheme; those get written down immediately because they are being remembered for the text itself, not the language. And pass three has a deadline: if a twice-met word takes more than two months to reappear, it dies in limbo rather than graduating, because a word that slow is not part of the living core of whatever I am reading.

The number I like, which is the reason this stayed: the sentences I wrote from graduated entries started appearing in my own writing unbidden, about one in three within a month. Under the old firehose method, the ratio was so close to zero I had stopped counting.

← all notes