Putting it in is not getting it back.

Endel Tulving spent the 1960s showing that a test trial can teach as much as a study trial. In 1967 he had people learn word lists under different mixes of presentation and recall. The surprise was that recalling the list was not a neutral measurement. It changed what they could recall later.

He also split availability from accessibility (Tulving and Pearlstone, 1966). A word can be in memory and still fail to come out until the right cue shows up. Encoding specificity (Tulving, 1973) is the same idea with a sharper edge: you retrieve best when the cues at test match the cues at study.

Henry Roediger and Jeffrey Karpicke brought that into classrooms. In Psychological Science (2006), students who practiced recalling a prose passage kept more of it a week later than students who reread it. Rereading felt better at five minutes. Retrieval won at two days and at a week. In Science (2008) they showed that once you can produce an item, more study of it does almost nothing, while more retrieval of it does a lot.

Inlet Language treats those as two dials. Tap-to-define is an encoding aid: meaning and sound sit on the word so you do not leave the page. The check after a text, and the due-word count on home, are retrieval. The feed is a new page, which is retrieval in a new cue set, closer to how you will meet the word in the wild.

If we only helped you look words up, we would be a glossary. If we only quizzed you, we would never have given the word a context to encode. The reader has to do both, and it has to know which one this minute is for.

Cognitive load is why you cannot retrieve thirty new words in one sitting. The unknown-word band is how we keep encoding from drowning retrieval.