Where does levels of processing collide?

Craik and Lockhart proposed in 1972 that how well something is remembered depends on the depth at which it is processed when first encountered, not on how long it sits in a short-term store or whether it transfers to a long-term one. Shallow processing attends to surface features such as how a word looks or sounds; deep processing attends to meaning, and deep processing was said to leave a more durable trace. Craik and Tulving's 1975 experiments seemed to bear this out: words judged for meaning were later recognised far better than words judged for their case or their rhyme. The framework was offered as an alternative to the multi-store model of memory, moving the explanation from where information is held to what the mind does with it. Here depth is the load-bearing word, and the rest of this page is about whether it was ever pinned down.[1][2]

Unit

A person's memory for material, as a function of how that material was processed at encoding

Recovery means

Not applicable

Measured by

Later recall or recognition after an encoding task that directs attention to surface features or to meaning

The trouble Baddeley pointed out in 1978 is that depth was never given a definition independent of the memory it was meant to explain. In practice a task counted as deep because it produced good retention, and it produced good retention because it was deep, so the claim folded back on itself and did little predictive work. There was no ruler for depth: no way to say in advance that one encoding task lay deeper than another, except by running the very memory test the theory was supposed to predict. Craik himself granted in a 2002 retrospective that the framework described an effect more than it explained one, and that depth had stayed frustratingly hard to measure on its own terms. This is a construct-validity fault about a core term that carries the whole account yet was never fixed operationally.[3][4]

Unit

The definition of depth, judged against the very memory scores it is used to explain

Recovery means

Not applicable

Measured by

Not directly measurable, which is the complaint: no index of depth exists apart from the retention it is meant to predict

Morris, Bransford, and Franks offered a different reading of the same results in 1977. On their account what helps memory is not depth in the abstract but the match between the mental operations used at encoding and those the retrieval test later demands: a meaning-based encoding wins on a standard recognition test, but a rhyme-based encoding wins when the later test itself asks about rhyme. In their experiments a shallow, sound-based encoding beat a deep, meaning-based one once the retrieval test rewarded sound, an outcome the depth story cannot accommodate. The effect Craik and Lockhart described is real, but it may be a property of the fit between study and test rather than of any fixed ladder of processing depth. This does not name a wholly new thing so much as relocate the old one, moving the explanation from depth to correspondence.[5]

Unit

A person's memory for material, as a function of how well encoding operations match retrieval demands

Recovery means

Not applicable

Measured by

Recall or recognition where the retrieval test is varied to reward the same feature the encoding task emphasised