New candidate diagnostics for cognitive decline and dementia have recently been proposed based on effects such as primacy and recency in word learning memory list tests. The diagnostic value is, however, currently limited by the multiple ways in which raw scores, and in particular these serial position effects (SPE), have been defined and analyzed to date. In this work, we build on previous analyses taking a metrological approach to the 10-item word learning list. We show i) how the variation in task difficulty reduces successively for trials 2 and 3, ii) how SPE change with repeated trials as predicted with our entropy-based theory, and iii) how possibilities to separate cohort members according to cognitive health status are limited. These findings mainly depend on the test design itself: A test with only 10 words, where SPE do not dominate over trials, requires more challenging words to increase the variation in task difficulty, and in turn to challenge the test persons. The work is novel and also contributes to the endeavour to develop for more consistent ways of defining and analyzing memory task difficulty, and in turn opens up for more practical and accurate measurement in clinical practice, research and trials.