<aside> ⚠️ I had this drafted for a while and decided to publish just the descriptive summary. Analysis & Conclusion TBD.

</aside>

<aside> 🪝 A synopsis of the literature, methodology & self-assessment of using Digit Span to quantify & evaluate short-term memory.

</aside>

Abstract

The philosophical construct of a finite span of memory dates back to Leibniz; he suggested that individuals have a finite capacity to process or hold information. He termed this capacity the span of apperception [1]. Digit Span is a psychological test to assess working memory and/or the ability to cache a circumscribed amount of information for a short period of time.

The classical methodology accrues to Wechsler [2], when he included Digit Span in the Wechsler Adult Intelligence Scale (WAIS) – an IQ test designed to measure intelligence and cognitive ability in adults. The Digit Span test has been shown to load on verbal comprehension, working memory, attention & information processing [3, 4, 5].

Experiment

Methodology

<aside> 📜 I’ve followed the experimental design of this paper, which improves upon the original Wechsler test.

</aside>

The subject is shown an increasing length of list of digits (e.g. {1, 4}, {5, 1, 3}). For each list, the digits are flashed sequentially with a delay of one second b/w each digit. The experiment follows a 1:2 staircase approach; a single correct response increases the length of the subsequent list by one digit, while two consecutive incorrect responses decrease the subsequent list length by one.

In Forward Span (FS), the participant responds with the recalled order of the digit list. In Backward Span (BS), the participant responds in reverse order of the original list.

Forward Span demo. I wrote a Python script to flash random digit list (of a given length) and assert my memorised response (for FS & BS).

Each test (FS & BS) carries till 15 trials. The entire experiment is repeated after a random period (b/w 15-20 days). A total of three experiments are carried out to minimise any dispositional errors such as fatigue, distraction, etc.

Experiment log

Scoring

Four measures were tracked for each experiment. Two of these (TE-TL & TE-ML) can be used for baseline comparison with the original test per WMS-III. MS is shown to be the most effective measure [6].

Table 1: a sample of one experiment (BS).

Two error maximum list length (TE-ML) Measures the maximum list length prior to the first two consecutive misses. In Table 1, the TE-ML score is 10 (reached length 10, prior to two consecutive misses at length 11).