Decisions are often biased by implicit or explicit preferences for one response alternative over another. Recent work suggests that such decision biases are influenced by task-evoked (phasic) increases in neuromodulation, as measured by pupil size, in line with theories proposing that neuromodulatory systems may regulate the balance between bottom-up and top-down processes. Although intriguing, there is scarce causal evidence tying neuromodulator activity to decision biases. Here, we manipulated decision bias through non-monetary punishments to specific errors (false alarms or misses) while upregulating baseline neuromodulator activity (through pharmacology, targeting two different neuromodulatory systems) in humans performing a visual detection task. Additionally, neural activity was measured with electroencephalography (EEG) and trial-by-trial fluctuations in baseline and task-evoked neuromodulator activity with pupil size. As anticipated, participants were more conservative in reporting target presence when false alarms were punished and more liberal when misses were punished. Computational modeling of choice behavior and response-locked parietal activity revealed bias-driven accumulation of sensory evidence over time towards a decision threshold. Crucially, decision bias was not related to elevated levels of baseline neuromodulation, neither when pharmacologically induced nor when measured through pupillometry. In contrast, we consolidate and extend the previously observed relationship between task-evoked neuromodulation (pupil size) and decision bias. Specifically, responses violating the dominant decision bias were associated with enlarged pupils preceding the response, which was reflected in parietal accumulation signals as well. Overall, we show that task-evoked (phasic), but not baseline (tonic), neuromodulator activity regulates how sensory evidence is accumulated towards a threshold during biased decision-making.