Action framing modulates the Pavlovian bias in go and no-go learning

Reward- and punishment-associated (i.e., Pavlovian) cues can trigger specific responses that interfere with learning appropriate actions. For instance, several studies have shown that reward impedes the learning of no-go responses, while punishment impedes the learning of go responses. This so-called Pavlovian bias is theorized to arise at the motor level, and to reflect a tight or even “hardwired” coupling between reward and action activation, and between punishment and action suppression. Contrary to this view, across four experiments testing adults (one pre-registered, total N = 585), using behavioral analyses and computational modeling, we show that action framing strongly modulates this Pavlovian bias. When go is framed as approach and no-go as avoidance, we replicated the typical asymmetry in learning. However, when go is framed as avoidance and no-go as approach, this Pavlovian bias disappeared. The Pavlovian bias thus operates partly on a semantic level, and is less hardwired than often theorized.