Les recherches présentées par Syssau et Monnier (2012) s’inscrivent dans une longue tradition de travaux sur les liens entre mémoire et émotion, en psychologie normale et pathologique. La valence émotionnelle a été prise en compte dans le paradigme DRM pour documenter l’amnésie traumatique. Élaborée par Pierre Janet sous sa forme moderne, l’amnésie traumatique (dissociative) est explorée en Europe et outre-Atlantique dès la fin du xix^e siècle. Ces anciens modèles, désignés de « dynamiques » car centrés sur la force du sujet, donnaient à l’activité un important rôle causal sur les émotions et la mémoire. Or, la prise en compte d’une variable dynamique (action) permet d’éclairer certaines difficultés des résultats expérimentaux récents, interprétés majoritairement à l’aide de facteurs sémantiques. Les modèles psychodynamiques, en particulier celui de Janet, conservent donc tout leur intérêt pour la recherche contemporaine.

This research by Syssau and Monnier (2012) pertains to a long tradition of works about the links between memory and emotion, for normal as well as for pathological psychology. Emotional valence was imported into the DRM paradigm in order to investigate traumatic amnesia. Elaborated by Pierre Janet in its modern formalization, traumatic (or dissociative) amnesia was investigated as soon as in the beginning of the 19th century in Europe and overseas. These former models, called “dynamic” because of their focus on the subject's force, gave to action an important causal role on both emotions and memory. Taking a dynamic variable (action) into consideration may help shedding light on some difficulties encountered by recent experimental results, which are generally interpreted within a semantic frame. Drawing upon Ribot's law (retrograde amnesia), Janet already noted what is now called an emotional enhancement of memory (EEM). Because contemporary interpretations fail to explain the facts entirely, Janet's ones in terms of force/fatigue of the subjects are still worth the reading: in what could be called the Ribot-Janet's law, the gradient of the emotional valence of words in mixed lists should be considered prior to the value of the emotional valence. One possible argument is that with mixed lists, the task is closer to a comparison of the emotional valence that to an assessment of it: an aspect of Janet's psychodynamic model too often neglected though still relevant for current psychological research. What's more, according to Janet's “perceptive scheme” (the closest today equivalent being Gibson's affordance), words denoting objects could trigger the action characteristic of the object. This is particularly relevant with manipulable objects, especially selected to be concrete for young children. Therefore, the items might not be coded only on a semantic network but also as motor and planning scripts: because no activity factor is considered in the current models, this could account for the difference observed between the predictions and the results, especially those regarding the false recognitions. Experimental psychodynamic models, Janet's ones in particular, have not yet proven false, and deserve to be reassessed within contemporary psychological research, where they are still thought-provoking.