This study explored how pretrial publicity (PTP) exposure and jury composition affect the slant and content of jury deliberations. It compares jury deliberations (N = 63) composed of jurors (N = 333) who are exposed to the same type of PTP (pure juries: all exposed to negative-victim [NV], negative-defendant [ND], or no PTP) with juries composed of jurors exposed to different types of PTP (mixed juries, e.g., half exposed to ND PTP and half to no PTP). Jury composition was found to bias the slant of trial evidence discussion, with pure-NV juries demonstrating a prodefense bias and pure-ND juries a proprosecution bias. This bias was also evident on mixed juries consisting of PTP-exposed jurors and no-PTP jurors. Importantly, biased evidence discussion mediated the effect jury composition on postdeliberation guilt assessments and was responsible for the spread of PTP bias from ND-PTP jurors to no-PTP jurors during deliberations. In addition, the deliberations of pure-NV juries showed evidence of being verdict driven by spending less time discussing trial evidence and taking more frequent straw polls than pure no-PTP juries. Finally, jury composition influenced the frequency of PTP discussion (pure ND-PTP juries discussed more frequently than pure NV-PTP juries) but not the frequency of PTP correction nor its effectiveness. Implications include that jury deliberations are not wholly effective at reducing PTP bias, regardless of jury composition. This is thought to be attributable to jurors’ biased encoding of trial evidence and source misattributions, resulting in biased trial evidence discussion and PTP discussion during deliberations.