Generating natural sounding synthetic speech from text requires a division of a text into IPs and assigning pauses between those phrases. A difficulty which faces attempts to model pauses quantitatively is high degree of variability exhibited by speakers in pause placement and duration. The present study seeks to investigate if Synchronous Speech (speech elicited when two speakers are asked to read a text together) can be used as a mean to reduce inter-speaker variability providing more reliable data for accurate modeling pause durations at IP breaks. We find reduced variability in pause duration when speakers read a text in synchrony. We also find an apparent dependence of pause duration on the length and/or syntactic complexity of the preceding phrase. The reduction in variability when reading synchronously is most evident for the one pause exhibiting markedly longer mean duration.