Twenty-six subjects performed orthographic, grammatical, evaluative, and self-referent orienting tasks. During each trial, integrated facial and forearm EMG activity, HR, and T-wave amplitude were monitored; following each trial, response latency and either reported task difficulty (Replication 1) or cognitive effort (Replication 2) were assessed; and recall was assessed at the conclusion of the session. Recall was poorest when words were judged in terms of their orthographic appearance, moderate when words were judged in terms of their grammatical or evaluative features, and best when words were judged for their self-descriptiveness, even though response latency was longest and reported effort was greatest for the grammatical task. Results also revealed that somatovisceral responses varied across tasks, with clear differences emerging between semantic and nonsemantic processing and between evaluative and self-referent processing. Results suggest that cognitive effort, rather than encoding efficacy, influences task-evoked somatic responses.