158 / 218
156
Chapter 7
loops with functionally similar cortical regions, when moving from the reward circuit to the
cognitive and subsequently action circuit (Haber, 2003; Haber et al., 2006; Haber and Knutson,
2010; Choi et al., 2012; Oh et al., 2014). We provide functional evidence that this gradient in
the striatum is indeed associated with task-related processing, where signaling in the anterior
caudate nucleus was associated with reward processing and increasingly posterior and lateral
portions of the striatum were involved in cognitive and motor processes.
Our observation that stimulation of the aPFC can modulate processing in distinct regions
of the striatum in a task-specific way provides evidence for the hypothesis that integration
of task-related information can occur across circuits, rather than being restricted within
parallel loops (Haber et al., 2000). Tracing work in non-human primates has revealed that the
dlPFC projects primarily to the dorsal caudate nucleus, but that its projection field extends
to other regions in the striatum, primarily the anterior and medial putamen. A similar
pattern was observed for cortical regions involved in reward processing: these projections
terminated primarily in the ventral striatum, but extent to more dorsal regions of the
striatum, including the caudate nucleus (Haber et al., 2006). Although this tracing work did
not include the anterior PFC, this work does suggest that these connections could underlie
the currently observed interaction between reward, task switching and response switching; a
functional, task-related gradient has been shown previously in the frontal cortex (for reviews
see: Koechlin and Summerfield, 2007; Botvinick, 2008; Badre and D’Esposito, 2009), where
increasingly anterior portions of the frontal cortex are involved in increasingly abstract goals.
In addition, using a computational modeling approach, Badre and Frank (2012) recently
suggested that a dopaminergic corticostriatal mechanism may underlie the integration
across rules when participants need to learn about task structure. We provide, for the first
time, functional evidence that integration across corticostriatal circuits takes place when
integrating information across functionally dissociable goals by showing that stimulation of
the aPFC can have functionally specific effects in distinct regions in the striatum depending
on the level of interaction (i.e. evidenced by a reward-related effect of aPFC stimulation in
the caudate nucleus, but in the putamen during the integration of reward, task switching and
response switching (
figure 7.5
).
However, the mechanisms underlying this integration across circuits are poorly understood.
We put forward three ways by which information about rewards can influence cognitive-
and action goals. The first involves direct cortico-cortical projections (Fuster, 2001; Wood
and Grafman, 2003; Badre and D’Esposito, 2009), while the second and third way assign
also a role to the striatum in the integration across corticostriatal circuits: via direct cortico-
striatal connections or via spiraling dopaminergic connections between the striatum and
the midbrain, i.e. striatal-nigral-striatal (SNS) connections (Haber et al., 2000; Haber, 2003;
Haber and Knutson, 2010). Future work should directly test these hypotheses, but here we
highlight in particular, based on recent convergent evidence from studies using the same
experimental paradigm, a role for the dopaminergic midbrain SNS projections in conveying
information about rewards to cognitive and action circuits. Specifically, the SNS account is