For decades, neuroscience and computational modeling have favored a clean, reductionist view of neuromodulation: dopamine signals reward prediction error, serotonin regulates mood, acetylcholine gates attention, and norepinephrine drives arousal. This "one chemical, one function" paradigm is elegant, but it fails to explain how the brain solves complex, multi-dimensional cognitive problems. In reality, these chemical systems do not operate in isolated silos. Instead, cognitive control emerges from the intricate interplay, sequential orchestration, and co-release of multiple neuromodulators.
Why Single Neuromodulators Fall Short
To understand why the brain relies on an interconnected neurochemical network, consider the challenge of reinforcement learning in a dynamic environment. A single scalar signal, like dopamine-mediated reward prediction error, can tell an agent whether an outcome was better or worse than expected. However, real-world decision-making requires tracking multiple variables simultaneously: positive payoffs versus negative costs, exploration versus exploitation, and confidence in one's internal model versus environmental volatility.
If dopamine alone handled these computations, the signal would become hopelessly conflated. By layering multiple neuromodulatory systems, the brain expands its computational degrees of freedom. For instance, Worbe et al. demonstrated that while dopamine heavily influences reward-based strategies, serotonin plays a valence-dependent role in cognitive control, specifically modulating model-based planning during loss avoidance.
Similarly, Rajabi et al. showed that norepinephrine and dopamine differentially modulate the meta-cognitive layer of decision-making. While blocking norepinephrine reduces perceptual self-awareness but boosts complex planning, boosting dopamine strengthens the link between a person's internal confidence and their subsequent strategic choices. By operating in tandem, these chemicals allow the brain to separate the evaluation of an action from the meta-control required to alter learning strategies.
Temporal and Functional Coordination: Opposing and Sequential Dynamics
How do these systems interact without causing chemical chaos? The brain employs two primary strategies: functional opposition and temporal sequencing.
In functional opposition, two neuromodulators act as a push-pull mechanism to balance behavioral states. A classic example is the relationship between appetitive and aversive processing. While dopamine drives approach behaviors toward rewards, computational frameworks by Huys et al. and Huys et al. suggest that serotonin acts as a counterweight, mediating behavioral inhibition and pruning decision trees to prevent the exploration of negative outcomes. Building on this, Wert-Carvajal et al. modeled how this direct dopamine-serotonin interplay gates synaptic plasticity in the hippocampus, with dopamine promoting spatial learning and serotonin inhibiting it to facilitate flexible reversal learning.
[High Acetylcholine] ---> Promotes Exploration & Synaptic Encoding
│
▼ (Time passes / Action taken)
[High Dopamine] ---> Gates Reward Delivery & Locks in Plasticity
Alternatively, neuromodulators can act in a precise temporal sequence to gate learning. Brzosko et al. discovered that the sequential activation of acetylcholine and dopamine is crucial for reward-based navigation. Acetylcholine first sets the cellular stage by enhancing excitability during exploration, and the subsequent arrival of dopamine locks in the synaptic weight changes required for successful navigation. This sequential gating allows the brain to solve the credit assignment problem, linking past exploratory actions to delayed rewards.
Local Interplay: Co-Transmission and Receptor Subtypes
The interaction between chemical systems is not merely a macro-level phenomenon driven by global brain states; it occurs locally at the level of individual synapses and microcircuits.
First, neurons frequently co-release classical neurotransmitters alongside neuromodulators. As Avvisati et al. highlighted, midbrain dopamine neurons do not just fire uniform dopamine signals; they exhibit highly diverse distributional coding and co-release glutamate and GABA. This co-transmission adds a layer of fast, localized synaptic signaling that complements the slower, volume-transmitted effects of dopamine. Even the chemical stability of these molecules is interdependent: Wu et al. uncovered a fundamental protective relationship where ambient glutamate prevents the neurotoxic oxidation of catecholamines like dopamine.
Second, the specific receptor isoforms expressed on target neurons dictate how these chemical signals are integrated. Within the striatum, Centonze et al. demonstrated that the short (D2S) and long (D2L) isoforms of the dopamine D2 receptor play distinct, segregated roles in modulating local glutamate and GABA transmission. This microcircuit-level segregation allows a single neuromodulator to precisely fine-tune the balance of excitation and inhibition in response to broader network demands.
Moving Beyond Single-System Models
Despite the clear computational necessity of multi-system interactions, historical biases have left significant gaps in our understanding. A scoping review by Severino et al. revealed that empirical research on "molecular connectivity"—how different brain regions interact via neurochemical systems—remains heavily skewed toward dopamine and serotonin in the context of Parkinson's disease, leaving many other systems under-explored.
Early computational frameworks often mirrored this isolation. For example, the classic septohippocampal model by Myers et al. focused strictly on cholinergic disruptions in associative learning, and Möller noted that standard basal ganglia models rely on highly simplified representations of dopamine dynamics that ignore its complex temporal interactions with other chemicals.
The frontier of the field lies in integrating these disparate pieces. When models expand to include multiple chemical actors, they unlock sophisticated cognitive phenomena. For instance, Prince et al. demonstrated that modeling the joint but separable actions of acetylcholine and noradrenaline explains how the hippocampus balances memory generalization (via acetylcholine) with memory discrimination (via noradrenaline). By treating the brain's neurochemical landscape as a coordinated, multi-layered network rather than a collection of independent channels, researchers are finally uncovering the true chemical architecture of cognitive control.
Key Foundational Work
A Computational Model of Cholinergic Disruption of Septohippocampal Activity in Classical Eyeblink Conditioning (1996) — contributed 1 concept elements to this direction.
Differential contribution of dopamine D2S and D2L receptors in the modulation of glutamate and GABA transmission in the striatum (2004) — contributed 1 concept elements to this direction.