Humanbenchmark.com includes many tests, the first of which is called Sequence Memory. You can try the test here.
The way it works is nine squares appear on the screen and then some of them light up in sequential order. At level one, one square lights up. At level two, two squares light up. Level three, three squares light up. etc. Your job is to wait for the sequence to finish and then click on the squares in the same sequence they lit up. The test progresses in difficulty until you make a mistake.
It’s interesting to note that when the Wechsler intelligence scale for children (WISC) was first revised (WISC-R) Wechsler tried to add a subtest much like this. In the 1974 WISC-R manual he writes:
All twelve tests were administered to the entire standardization sample, as was also a thirteenth test entitled Visual Motor Memory. The latter test, requiring the child to copy a sequence tapped by the examiner, was intended as a supplementary Performance test and as a non-verbal visual motor analog of Digit Span. In spite of certain merits, the test was eliminated from the WISC-R because it posed problems of administration and failed to meet some of the rigorous statistical standards that were applied.WISC-R manual page 8
I can see how it would pose administration problems because the examiner would need an excellent visual memory herself just to see if the child repeated the sequence correctly, especially if he did so quickly. But when the test is administered by computer like on humanbenchmark.com, that’s no longer an issue.
Nonetheless, the newest edition of the WISC (WISC-V) includes subtests like Picture Span and Spatial Span. In Picture Span, the child sees a bunch of pictures in correct order and then must remember their sequence when present in random order. This draws on similar cognitive skills as humanbenchmark’s sequence memory test and it correlates 0.61 with full-scale IQ (a strong proxy for g) among 16-year-olds. Oddly, the spatial span subtest (where a child must tap a bunch of blocks in the same sequence he sees the examiner do it) only correlates 0.43 (WISC-V technical manual, page 69), even though this includes a backwards spatial span component. When it comes to auditory sequencing, going from backward to forward doubles the g loading so I’d love to know why picture span is so much more g loaded than spatial span, despite having no backward component. Maybe there’s too much error because the examiners can’t keep up?
Humanbenchmark.com provides percentiles for scores on the sequence memory test, but there’s no context. Does the percentile reflect how many people we beat or how many attempts we beat (many by the same person) ? And how self-selected is the reference sample? It would be interesting to know how my readers (mean IQ 130) do on this test, how their scores respond to practice, and how this relates to their scores on established psychometric tests.