Sternberg’s paradigm is currently viewed as a typical short-term memory task and is widely used to tap mnemonic capacities in neuroscience studies. However, Sternberg’s original procedure includes an experimental constraint – recalling the sequence of digits in order – which was not reused in the following studies. In previous research (Corbin & Marquer, 2008, 2009), we showed that the recall constraint has an impact on the quantitative results as well as on the strategies implemented. These findings led us to wonder whether the presence or absence of this simple experimental constraint could also affect the processes implemented in Sternberg’s task. In order to answer this question, we analyzed the relationships between the performance levels of 50 participants on Sternberg’s task on various well-known span tasks and on a classical visual search task. The results showed that, in the recall condition, Sternberg’s paradigm appears to be a verbal working memory task, whereas in the no-recall condition, the task appears to be a recognition task that involves visuospatial memory capacities. In this latter condition, the processes implemented may be more similar to those implemented in visual search tasks.