It has been widely reported that intrinsic brain activity, in a variety of animals including humans, is spatiotemporally structured. Specifically, propagated slow activity has been repeatedly demonstrated in animals. In human resting-state fMRI, spontaneous activity has been understood predominantly in terms of zero-lag temporal synchrony within widely distributed functional systems (resting-state networks). Here, we use resting-state fMRI from 1,376 normal, young adults to demonstrate that multiple, highly reproducible, temporal sequences of propagated activity, which we term “lag threads,” are present in the brain. Moreover, this propagated activity is largely unidirectional within conventionally understood resting-state networks. Modeling experiments show that resting-state networks naturally emerge as a consequence of shared patterns of propagation. An implication of these results is that common physiologic mechanisms may underlie spontaneous activity as imaged with fMRI in humans and slowly propagated activity as studied in animals. Significance It is well known that slow intrinsic activity, as measured by resting-state fMRI in a variety of animals including humans, is organized into temporally synchronous networks. The question of whether intrinsic activity contains reproducible temporal sequences has received far less attention. We have previously shown that human resting-state fMRI contains a highly reproducible lag structure. Here, we demonstrate that this lag structure is of high dimensionality and consists of multiple highly reproducible temporal sequences, which we term “lag threads.” Moreover, we demonstrate that the well-known zero-lag temporal correlation structure of intrinsic activity emerges as a consequence of lag structure. Thus, lag threads may represent a fundamental and previously unsuspected level of organization in resting-state activity.