Traditionally, dreams have been seen as experiences that one cannot control, as something that happensto the dreamer (at times through involvement of supernatural powers), without the dreamer’s permission,volition or agency. This view was famously challenged in the advent of psychoanalysis: in his Interpretation ofDreams, Freud proposed that while we may not be consciously in control of our dreams, our unconscious mindis actively constructing dream content, and that dream content is symptomatic of our repressed, accumulatedneuroses (Freud, 1900/2010). This shift in perspective signaled that dreams may in fact be subject to individualexperience and to one’s mental state (which is something that is possible to change, or even to control),therefore bringing, at least partially, responsibility for dreams to the dreamers themselves. Emergingneuroscience of dreams, in 1970s, however, adopted a more conservative behaviorist position, and for arelatively long time the dominant view has been one of neuro-reductionism, where dreams were seen as randomhallucinatory products of the activity of the sleeping brain (Hobson & McCarley, 1977). In recent years, a more2nuanced picture of dreams is gradually emerging. Research from psychology, philosophy and anthropologyconverges on the idea that dreams may be individual or even collective practices rather than uncontrolled brainevents. Developmentally and temporally, dreaming can be recognized as cognitive achievement (alongside othercognitive abilities, such as memory, perception, attention, etc.) (Foulkes, 2014). And dream qualities, includingwhat is possible in the dream state, how rich the dream experience is, and how well the dream will beremembered, may change as a result of attentional practices during waking hours. Research on dreamincubation, dream sharing and lucid dreams shows that the dreamer is an active participant and co-creator oftheir dream life, and that the dreamer’s agency, awareness and degrees of control are all dynamic, continuousand potentially trainable skills. Further, in line with work on 4E cognition (Menary, 2010) and followingevidence from sensory incorporation studies (Nielsen, 1993, 2017; Sauvageau, Nielsen, & Montplaisir, 1998), ithas been proposed that dreams are not simply experiences of virtual reality confined in the sleeping brain, butrather can be conceptualized as processes of embodied imagination (Thompson, 2014; Solomonova & Sha,2016; Solomonova, 2017), rooted in lived sensorimotor experience and responsive to sensory information fromthe outside world.In this chapter, we review the different ways that attention works in relation to dreams and how it mayfunction in dreams, and apply the framework of attention, proposed in this volume – as a means of accessingand mediating interactions with the world - to the dreaming world. We first review prior work on the role ofattention as 1) access to dreams, e.g., how practices of recording and sharing dreams act as enabling factors forimproving dream recall and enhancing richness of dream experience; and as 2) a mediator of dreams, e.g., howincubation, imagery rehearsal, and ultimately lucidity can be cultivated as cognitive skills enabling agency in thedream experience. We propose that attention functions as a constitutive factor in dream experience and that it isa trainable, developmental cognitive skill. We argue that dreams are not simply experiences that happen to thedreamer, rather, through employing attentional techniques in various ways, the dreamer may cultivate differentdegrees of agency in the dream.