Abstract of talk presented at the Toward a Science of Consciousness Conference IV, Tucson, April 10, 2000

Smooth Tracking Eye-Movements Discriminate Both Dreaming And Perception From Imagination

S. LaBerge and P.G. Zimbardo
(Department of Psychology, Stanford University)

In the minds of many scientists, dreaming seems more closely related to imagination than to perception. For example, Foulkes (1985) explicitly denies that dreaming is in any form related to perception, claiming instead that it is a "symbolic act" more like thinking. In contrast, we have argued that the only necessary difference between waking "perceptions" and dreaming "hallucinations" is the presence or absence, respectively, of external sensory input. In our view, dreaming and perception are two special cases in which the contents of consciousness are respectively independent of, or constrained by sensory input (LaBerge, 1998, Llinas & Pare, 1997, etc).

In order to directly compare dreaming, perception, and imagination we designed a visual eye movement tracking task that could be carried out analogously in all three states. The eye movement tracking task was also selected as a test of the "scanning hypothesis", which hypothesizes that the subjective gaze shifts of the dreamer looking around the dream are accompanied by actual eye movements and fixations in the corresponding directions and locations.

Four men and two women ranging in age from 20-48 years participated in the study. All had excellent dream recall (typically remembering more than one dream per night) and had received training in lucid dreaming (LaBerge & Rheingold, 1990) so that they were able to recognize when they were dreaming and carry out pre-agreed upon actions in their dreams. They spent between 1 and 8 nights in the laboratory during the course of the study.

On each night, subjects were fitted with es to measure EEG (29 channels), sub-mental EMG, and vertical and horizontal EOG (DC-50Hz). Physiological data were saved to disk using a digital amplifier.

Subjects were instructed to signal when they realized they were dreaming by means of a pair of left-right eye movement signals (LRLR) (LaBerge, et al., 1981). Then they were to carry out the tracking protocol: follow the movement of the tip of the right index finger as it moves clockwise in a circle centered in the visual field; signal with LRLR; track a circle counterclockwise; and signal with LRLR. The tracking protocol was also carried out while awake with eyes open ("perception") and eyes closed ("imagination") before and after sleep.

[Circle tracking: Perception, Dreaming, and Imagination]

Figure 1. Circle tracking compared in three states.

The subjects succeeded in becoming lucid and carrying out the tracking task in their dreams a total of 14 times. EOG data were adjusted according to pre- and post-sleep calibrations and computed angular velocities were converted to standard scores (EMVZ). Mean values of EMVZ were significantly higher (p<.0001) during the imagination condition (+0.95, SD=0.97) than during both perception (-0.66, SD=0.26) and dreaming (-0.54, SD=0.30), reflecting the fact that imagination was much more likely to be accompanied by saccadic eye movements, while dreaming and perception showed predominately slow tracking eye movements. Mean values of EMVZ for perception and dreaming were not significantly different. See figure 2.

[Mean eye-tracking velocity by state]

Figure 2. Mean values of standardized eye-movement velocities (EMVZ) as a function of state. The grey boxes contain 50% of the data, with median values, 75th and 25th percentiles marked.

A discriminant function using EMVZ to classify perception and imagination data correctly classified 85% of imagination cases and 100% of perception cases. The same function applied to the dreaming data classified 100% of the cases as perception. In other words, using mean eye movement velocity alone, we were able to correctly classify 95% of the cases, strongly supporting our hypothesis that as far as the visual vividness dimension is concerned, dreaming consciousness is nearly identical to waking perceptual consciousness, and just as distinct from imagination as imagination is distinct from perception.

The results provide strong support for the following form of the "scanning hypothesis": that shifts in the dreamer's gaze are accompanied by corresponding movements of the sleeper's eyes. They also support the isomorphism principle, although the direction of causality is opposite of that assumed by the Activation Synthesis model (Hobson & McCarley, 1977).

Foulkes (1985)
Hobson & McCarley, (1977)
LaBerge (1996)
LaBerge & Rheingold, H. (1990)
LaBerge, et al. (1981)
Llinas & Pare