Cet article propose une synthèse des travaux ayant eu recours à la tâche de décision lexicale avec amorçage sémantique chez les patients schizophrènes. Les résultats obtenus à l’aide de ce paradigme sont hétérogènes : les schizophrènes obtiennent des effets d’amorçage sémantique soit supérieurs, soit inférieurs, soit égaux à ceux des sujets sains. Ces résultats sont susceptibles de refléter, chez ces patients, un dysfonctionnement des processus automatiques et/ou contrôlés sous-tendant l’obtention de ces effets. La littérature montre que si un dysfonctionnement des processus contrôlés est largement étayé chez les patients schizophrènes, l’efficience du processus automatique de diffusion de l’activation reste controversée. Cette absence de consensus quant au fonctionnement du processus automatique peut être reliée aux manipulations des variables expérimentales (SOA, proportion de paires de mots reliés, consignes) effectuées par les auteurs, qui ne permettent pas d’évaluer avec certitude son efficience. D’autre part, certaines relations sémantiques privilégiées par les auteurs (amorçage par médiation) ne semblent pas être les plus pertinentes pour permettre de comprendre comment se diffuse l’activation sémantique chez ces patients. Enfin, certains sous-types de patients ont particulièrement été étudiés, rendant ainsi les généralisations à la population difficiles.

This article is a review of studies using the semantic priming paradigm to assess the functioning of semantic memory in schizophrenic patients.

Semantic priming describes the phenomenon of increasing the speed with which a string of letters (the target) is recognized as a word (lexical decision task) by presenting to the subject a semantically related word (the prime) prior to the appearance of the target word. This semantic priming is linked to both automatic and controlled processes depending on experimental conditions (stimulus onset asynchrony (SOA), percentage of related words and explicit memory instructions). Automatic process observed with short SOA, low related word percentage and instructions asking only to process the target, could be linked to the « automatic spreading activation » through the semantic network. Controlled processes involve « semantic matching » (the number of related and unrelated pairs influences the subjects decision) and « expectancy » (the prime leads the subject to generate an expectancy set of potential target to the prime). These processes can be observed whatever the SOA for the former and with long SOA for the later, but both with only high related word percentage and explicit memory instructions.

Studies evaluating semantic priming effects in schizophrenia show conflicting results : schizophrenic patients can present hyperpriming (semantic priming effect is larger in patients than in controls), hypopriming (semantic priming effect is lower in patients than in controls) or equal semantic priming effects compared to control subjects.

These results could be associated to a global impairment of controlled processes in schizophrenia, essentially to a dysfunction of semantic matching process. On the other hand, efficiency of semantic automatic spreading activation process is controversial. These discrepancies could be linked to the different experimental conditions used (duration of SOA, proportion of related pairs and instructions), which influence on the degree of involvement of controlled processes and therefore prevent to really assess its functioning. In addition, manipulations of the relation between prime and target (semantic distance, type of semantic relation and strength of semantic relation) seem to influence reaction times. However, the relation between prime and target (mediated priming) frequently used could not be the most relevant relation to understand the way of spreading of activation in semantic network in patients with schizophrenia. Finally, patients with formal thought disorders present particularly high priming effects relative to controls.

These abnormal semantic priming effects could reflect a dysfunction of automatic spreading activation process and consequently an exaggerated diffusion of activation in the semantic network. In the future, the inclusion of different groups schizophrenic subjects could allow us to determine whether semantic memory disorders are pathognomonic or specific of a particular group of patients with schizophrenia.