Constructing value in the medial prefrontal cortex

Thesis

Humans and other animals are remarkably adept at integrating a vast array of information into their choices. For instance, they seamlessly integrate their own internal states and goals (e.g. being thirsty or aspiring to obtain a PhD) into their actions without extensive training. How does the brain enable them to do this efficiently? In this thesis, I will investigate three key components of this behaviour: i) how humans track accumulation of resources as information about the current state of their goals, ii) how they arbitrate between two competing goals, and iii) how remembered value information is efficiently re-weighted to support pursuit of novel goals. All reported experiments employed functional magnetic resonance imaging to localize neural responses.

In the first line of research, covering points i) and ii), I will review two experiments, which investigate the neural correlates in the prefrontal cortex (PFC) during goal-directed decision-making. In Chapter 3, I report results from a study in which participants accumulated rewards from risky gambles over time. A key value-related area in the prefrontal cortex (PFC), the ventromedial PFC, encoded both the momentary receipt of reward and the participants’ current cumulative rewards. Next, in Chapter 4, I ask how humans arbitrate between two competing goals based on such latent accumulation signals. Another PFC region, the anterior cingulate cortex (ACC), reflected both the pressure to act on one goal over another as well as participants’ likelihood to act according to this pressure. Importantly, the vmPFC encoded how much participants’ choices redressed the imbalance between goals, indicating that its value correlates may represent an integration of multiple goals and momentary receipt. If goals exert pressure to re-direct choices towards certain courses of action, then the areas supporting such decision-making must be able to flexibly re-weight information according to these new goals. Thus, in Chapter 5, I present a study showing that the frontopolar cortex as well as the ACC are key areas supporting this dynamic re-weighting and the representation of opportunity costs. Lastly, I will review a framework of primary rewards, homeostatic reinforcement learning, and how it may be adapted to encompass cognitive goals.

Authors: Juchems, K

• DOI: 10.5287/ora-qryge400r
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English