It involves mimicking a researcher as they tap a sequence of up to nine identical spatially separated blocks.

The sequence starts out simple, usually using two blocks, but becomes more complex until the subject's performance suffers.

An fMRI study involving subjects undergoing this test revealed that while the sequence length increases, general brain activity remains the same.

Corsi blocks tasks with a normal forward order requires support from the visuospatial sketch pad, but not from the phonological loop.

In a study with visuospatial learning disabled (VSLD) children, they found that only the VSLD children had significantly impaired performance on the Corsi backward task as compared to the forward, while both the control group and the VSLD group showed poor performance on the digit span backward compared to the forward.