Objectives: This study aimed to evaluate the informational component of speech-on-speech masking. Speech perception in the presence of a competing talker involves not only informational masking, but a number of masking processes involving interaction of masker and target energy in the auditory periphery. Such peripherally generated masking can be eliminated by presenting the target and masker in opposite ears (dichotically). However, this also reduces informational masking by providing listeners with lateralization cues that support spatial release from masking. In tonal sequences, informational masking can be isolated by rapidly switching the lateralization of dichotic target and masker streams across the ears, presumably producing ambiguous spatial percepts that interfere with spatial release from masking. However, it is not clear if this technique works with speech materials. Design: Speech reception thresholds (SRTs) were measured in 17 young normal-hearing adults for sentences produced by a female talker in the presence of a competing male talker under three different conditions: diotic (target and masker in both ears), dichotic, and dichotic but switching the target and masker streams across the ears. Because switching rate and signal coherence were expected to influence the amount of IM observed, these two factors varied across conditions. When switches occurred, they were either at word boundaries or periodically (every 116 ms) and either with or without a brief gap (84 ms) at every switch point. In addition, SRTs were measured in a quiet condition to rule out audibility as a limiting factor. Results: SRTs were poorer for the four switching dichotic conditions than for the non-switching dichotic condition, but better than for the diotic condition. Periodic switches without gaps resulted in the worst SRTs compared to the other switch conditions, thus maximizing informational masking. Conclusions: These findings suggest that periodically switching the target and masker streams across the ears (without gaps) was the most efficient in disrupting spatial release from masking. Thus, this approach can be used in experiments that seek a relatively pure measure of informational masking, and could be readily extended to translational research.