Three experiments are reported that examine the masking effect on the detection of kinesthetic targets from noise presented simultaneously on the same hand or on a different hand as the target. Performance was facilitated when noise on the two hands was correlated versus uncorrelated or when it occurred on just one hand. The data are consistent with correlated input on the two hands being compared and used to reduce noise effects. Moreover, participants ignored uncorrelated noise from nontarget hands when they knew in which hand the targets would occur. There was no effect of shifts from uncorrelated to correlated noise 200 msec prior to targets, suggesting that such bottom-up changes were insufficient to induce comparison across the hands. In contrast, shifts from correlated to uncorrelated noise 200 msec prior to the target had performance at the same level as that with uncorrelated noise throughout the trial. This result indicates that the loss of correlation was detected within this interval and thatwhen the target's location was known, participants were then able to switch attention to the target hand. The interactions between attentional filtering and the correlation on the processing of bilateral kinesthetic inputs are discussed.