Search Papers | Poster Sessions | All Posters
Poster B26 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Computational mechanisms of aversive generalisation in anxiety
Luianta Verra1 (), Bernhard Spitzer1, Nicolas W. Schuck2, Ondrej Zika1; 1Max Planck Institute for Human Development, Berlin, Germany, 2Universität Hamburg, Hamburg, Germany
Excessive generalisation of threat to similar stimuli is characteristic in anxiety. Such generalisation can arise from a failure to correctly identify the threatening stimulus or from the transfer of learned values to similar stimuli. Here we use computational modelling to characterise how perceptual and value-based mechanisms shape generalisation functions and how they relate to anxiety. Participants (n=140) learned probabilistic stimulus-outcome associations that were then tested for generalisation on morphs of the original stimulus. Within each participant, we varied stimulus discriminability (high/low; perceptual manipulation) and the rate of reinforcement (25/50/75%; value manipulation). We found that participants generalized threat expectancy to new stimuli. Interestingly, participants generalized either by extrapolating linearly (linear function) or by using a similarity-based strategy (gaussian function). Both perceptual uncertainty and reinforcement rate impacted generalisation. Value generalisation was mediated by the generalisation strategy while perceptual uncertainty increased generalisation independently of it. Anxiety was associated with stronger generalisation for stimuli further from the original stimulus, especially when reinforcement rate was high. This study characterises different mechanisms of aversive generalisation and contributes to our understanding of excessive generalisation in anxiety.
Keywords: fear generalisation function learning anxiety