In 1976 J. Allan Hobson and Robert McCarley proposed a new theory that changed dream research, challenging the previously held Freudian view of dreams as subconscious dreamily wishes to be interpreted. Activation synthesis theory asserts that the sensory experiences are fabricated by the cortex as a means of interpreting chaotic signals from the pons. They propose that in REM sleep, the ascending cholinergic PGO (ponto-geniculo-occipital) waves stimulate higher midbrain and forebrain cortical structures, producing rapid eye movements. The activated forebrain then synthesizes the dream out of this internally generated information. They assume that the same structures that induce REM sleep also generate sensory information.

Hobson's 1976 research suggested that the signals interpreted as dreams originated in the brain stem during REM sleep. However, research by Mark Solms suggests that dreams are generated in the forebrain, and that REM sleep and dreaming are not directly related. While working in the neurosurgery department at hospitals in Johannesburg and London, Solms had access to patients with various brain injuries. He began to question patients about their dreams and confirmed that patients with damage to the parietal lobe stopped dreaming; this finding was in line with Hobson's 1977 theory. However, Solms did not encounter cases of loss of dreaming with patients having brain stem damage. This observation forced him to question Hobson's prevailing theory which marked the brain stem as the source of the signals interpreted as dreams. Solms viewed the idea of dreaming as a function of many complex brain structures as validating Freudian dream theory, an idea that drew criticism from Hobson. In 1978, Solms, along with partners William Kauffman and Edward Nadar, undertook a series of traumatic-injury impact studies using several different species of primates, particularly howler monkeys, in order to disprove Hobson's postulation that the brain stem played a significant role in dream pathology. Unfortunately, Solms' experiments proved inconclusive, as the high mortality rate associated with using a hydraulic impact pin to artificially induce brain damage in test subjects meant that his final candidate pool was too small to satisfy the requirements of the scientific method.

Evolutionary psychologists believe that dreams serve some adaptive function for survival. Deirdre Barrett describes dreaming as simply "thinking in different biochemical state" and believes people continue to work on all the same problems—personal and objective—in that state. Her research finds that anything—math, musical composition, business dilemmas—may get solved during dreaming, but the two areas especially likely to help are 1) anything where vivid visualization contributes to the solution, whether in artistic design or invention of 3-D technological devices and 2) problem where the solution lies in "thinking outside the box"—i.e. the person is stuck because conventional wisdom on how to approach the problem is wrong. In a related theory, which Mark Blechner terms "Oneiric Darwinism," dreams are seen as creating new ideas through the generation of random thought mutations. Some of these may be rejected by the mind as useless, while others may be seen as valuable and retained. Finnish psychologist Antti Revonsuo posits that dreams have evolved for "threat simulation" exclusively.

Combining Hobson's activation synthesis hypothesis with Solms' findings, the continual-activation theory of dreaming presented by Jie Zhang proposes that dreaming is a result of brain activation and synthesis; at the same time, dreaming and REM sleep are controlled by different brain mechanisms. Zhang hypothesizes that the function of sleep is to process, encode and transfer the data from the short-term memory to the long-term memory, though there is not much evidence backing up this so-called "consolidation." NREM sleep processes the conscious-related memory (declarative memory), and REM sleep processes the unconscious related memory (procedural memory).

Zhang assumes that during REM sleep, the unconscious part of a brain is busy processing the procedural memory; meanwhile, the level of activation in the conscious part of the brain will descend to a very low level as the inputs from the sensory are basically disconnected. This will trigger the "continual-activation" mechanism to generate a data stream from the memory stores to flow through the conscious part of the brain. Zhang suggests that this pulse-like brain activation is the inducer of each dream. He proposes that, with the involvement of the brain associative thinking system, dreaming is, thereafter, self-maintained with the dreamer's own thinking until the next pulse of memory insertion. This explains why dreams have both characteristics of continuity (within a dream) and sudden changes (between two dreams).

Eugen Tarnow suggests that dreams are ever-present excitations of long-term memory, even during waking life. The strangeness of dreams is due to the format of long-term memory, reminiscent of Penfield & Rasmussen's findings that electrical excitations of the cortex give rise to experiences similar to dreams. During waking life an executive function interprets long-term memory consistent with reality checking. Tarnow's theory is a reworking of Freud's theory of dreams in which Freud's unconscious is replaced with the long-term memory system and Freud's "Dream Work" describes the structure of long-term memory.

Dreams of absent-minded transgression (DAMT) are dreams wherein the dreamer absentmindedly performs an action that he or she has been trying to stop (one classic example is of a quitting smoker having dreams of lighting a cigarette). Subjects who have had DAMT have reported waking with intense feelings of guilt. One study found a positive association between having these dreams and successfully stopping the behavior.

A 2001 study showed evidence that illogical locations, characters, and dream flow may help the brain strengthen the linking and consolidation of semantic memories. These conditions may occur because, during REM sleep, the flow of information between the hippocampus and neocortex is reduced. Increasing levels of the stress hormone cortisol late in sleep (often during REM sleep) cause this decreased communication. One stage of memory consolidation is the linking of distant but related memories. Payne and Nadal hypothesize that these memories are then consolidated into a smooth narrative, similar to a process that happens when memories are created under stress

Lucid dreaming is the conscious perception of one's state while dreaming. In this state a person usually has control over characters and the environment of the dream as well as the dreamer's own actions within the dream. The occurrence of lucid dreaming has been scientifically verified.

Oneironaut is a term sometimes used for those who lucidly dream.

Robert (1886), a physician from Hamburg, was the first who suggested that dreams are a need and that they have the function to erase (a) sensory impressions which were not fully worked up and (b) ideas which were not fully developed during the day. By the dream work incomplete material will be either removed or deepened and included into memory. Robert's ideas were cited repeatedly by Freud in his Traumdeutung. Hughlings Jackson (1911) viewed that sleep serves to sweep away unnecessary memories and connections from the day. This was revised in 1983 by Crick and Mitchison's "reverse learning" theory, which states that dreams are like the cleaning-up operations of computers when they are off-line, removing parasitic nodes and other "junk" from the mind during sleep. However, the opposite view that dreaming has an information handling, memory-consolidating function (Hennevin and Leconte, 1971) is also common. Dreams are a result of the spontaneous firings of neural patterns while the brain is undergoing memory consolidation during sleep.

Coutts hypothesizes that dreams modify and test mental schemas during sleep during a process he calls emotional selection, and that only schema modifications that appear emotionally adaptive during dream tests are selected for retention, while those that appear maladaptive are abandoned or further modified and tested. Alfred Adler suggested that dreams are often emotional preparations for solving problems, intoxicating an individual away from common sense toward private logic. The residual dream feelings may either reinforce or inhibit contemplated action.

During sleep the eyes are closed, so that the brain to some degree becomes isolated from the outside world. Moreover all signals from the senses (except olfaction) must pass through the thalamus before they reach the brain cortex, and during sleep thalamic activity is suppressed. This means that the brain mainly works with signals from itself. A well-known phenomenon in dynamical physical systems where the level of input and output from the system is low is that oscillation makes spontaneous resonance patterns to occur. Hence, dreams may be the simple consequence of neural oscillation.