Independent component analysis reveals the unity of cognitive control

Clémence Roger (1) , Christian G. Bénar (2) , Franck Vidal (1) , Thierry Hasbroucq (1) , Boris Burle (1)

1 LNC - Laboratoire de Neurosciences Cognitives [Marseille]
2 Epilepsies, Lesions Cerebrales et Systemes Neuraux de la Cognition

In reaction time tasks, when subjects commit an error, a negative wave peaking approximatively 70-100ms after the erroneous response is recorded with EEG. This negativity, called "Error (Related) Negativity" (Ne or ERN[1, 2]), is maximal fronto-centrally, above the anterior cingulate cortex and/or SMA and was first interpreted as reflecting an error detection mechanism. However, after Laplacian estimation, a similar component was later observed on correct trials [3]. If this component on correct trials were to be the same as the one observed on errors, this would put important constraints on computational models of cognitive control. To address this issue we used Independent Com- ponent Analysis (ICA) to evaluate whether a single component (in ICA terms) could account for the waves observed in both erroneous and correct trials. For all the participants, a single component that accounts for the waves observed in the three categories of trials was found. The localisation of the sources is consistent with a rostral-cingulate zone origin, where control mechanisms are likely implemented [4]. This novel use of ICA allowed us to conclude that the negativities observed on error and correct trials are reflecting the same physiological mechanism whose amplitude is modulated as function of the performance.

Mots clés: Learning; Neuron and plasticity; Neuronal models; Synaptic plasticity

Domaines: Neurosciences [q-bio.NC]; Neurosciences

Clémence Roger, Christian G. Bénar, Franck Vidal, Thierry Hasbroucq, Boris Burle. Independent component analysis reveals the unity of cognitive control. Deuxième conférence française de Neurosciences Computationnelles, "Neurocomp08", Oct 2008, Marseille, France. ⟨hal-00331606⟩