Brain mechanisms for perceptual and reward-related decision-making

Gustavo Deco, Edmund T. Rolls, Larissa Albantakis, Ranulfo Romo

Research output: Contribution to journalReview ArticleResearchpeer-review

109 Citations (Scopus)


Phenomenological models of decision-making, including the drift-diffusion and race models, are compared with mechanistic, biologically plausible models, such as integrate-and-fire attractor neuronal network models. The attractor network models show how decision confidence is an emergent property; and make testable predictions about the neural processes (including neuronal activity and fMRI signals) involved in decision-making which indicate that the medial prefrontal cortex is involved in reward value-based decision-making. Synaptic facilitation in these models can help to account for sequential vibrotactile decision-making, and for how postponed decision-related responses are made. The randomness in the neuronal spiking-related noise that makes the decision-making probabilistic is shown to be increased by the graded firing rate representations found in the brain, to be decreased by the diluted connectivity, and still to be significant in biologically large networks with thousands of synapses onto each neuron. The stability of these systems is shown to be influenced in different ways by glutamatergic and GABAergic efficacy, leading to a new field of dynamical neuropsychiatry with applications to understanding schizophrenia and obsessive-compulsive disorder. The noise in these systems is shown to be advantageous, and to apply to similar attractor networks involved in short-term memory, long-term memory, attention, and associative thought processes.

Original languageEnglish
Pages (from-to)194-213
Number of pages20
JournalProgress in Neurobiology
Publication statusPublished - Apr 2013
Externally publishedYes


  • Attractor network
  • Confidence
  • Decision making
  • Dynamical neuropsychiatry
  • Noise in the brain
  • Reward value decision-making
  • Vibrotactile decision-making

Cite this