Aspen Movie Map

The Aspen Movie Map was an interactive "surrogate travel" system built at the Massachusetts Institute of Technology's Architecture Machine Group (ArcMac) between 1978 and 1980. It let a user drive through the streets of Aspen, Colorado, on a touch-sensitive screen, using thousands of film frames shot at fixed distances along every street and stored on computer-controlled optical videodiscs. Andrew Lippman directed the project, with funding from the Cybernetics Technology Office of the US Defense Advanced Research Projects Agency (DARPA), and Michael Naimark was responsible for its cinematography.

Besides street-level travel, users could switch to aerial views and maps, toggle between fall and winter footage, touch a building to see its facade, and step "inside" through slide shows and short films. Robert Mohl's 1981 doctoral thesis on the system, subtitled "An Investigation of Spatial Learning in Virtual Environments", tested whether people who explored Aspen on the movie map learned the town's layout. Later writers describe the project as a precursor of Google Street View and other navigation technology.

Reviewed 4 October 2026. All claims checked against Mohl's 1981 MIT thesis, Naimark's 2006 Presence essay and project page, the cited paper abstracts and the CHM, MIT Technology Review and MIT Media Lab pages. About review dates.

How it worked

The system was organized around a model of the town's street network. In Mohl's description, a user sat in front of a touch-sensitive television screen showing roughly what a driver would see through the windshield. Touching a speed control started movement down the street, other buttons turned onto side streets or switched to a view out of the side window, and a map traced the route traveled so far. Touching the map zoomed into an area, and the map could be switched between a street map, a landmark map and an aerial photograph. Touching a building brought up a close-up of its facade, a "season knob" flipped between winter and summer versions, and a slide show behind the facade served as a short "visit" to the building. Mohl notes that the project used the terms "movie map" and "surrogate travel" almost interchangeably: "surrogate travel" for the system as a model of experience, "movie map" for the system as a model of spatial structure.

Lippman's paper at SIGGRAPH 1980 described the system as "an interactive, dynamic map" built with videodisc technology "to engage the user in a simulated 'drive' through an unfamiliar space". The user could control "the speed, route, angle of view and mode of presentation" and could access ancillary data stored "in the buildings or in locales in the environment". The system was also shown in ArcMac's "Media Room", where the traveler sat in an instrumented armchair to control speed and direction while touch screens showed map and aerial views.

History

Origins

In early 1978 ArcMac received one of the first 25 prototype videodisc players from the MCA Corporation, together with a contract to master three discs of its own. Naimark describes the optical videodisc of the time as able to store half an hour of analog video (54,000 frames) and two-channel audio, with random access to every frame under computer control. In spring 1978, MIT undergraduate Peter Clay, helped by graduate students Bob Mohl and Michael Naimark, "mapped" the hallways of MIT by wheeling a single-frame 16 mm camera down them and shooting one frame per step. The footage went on the B-side of MIT's first laserdisc, and a simple program let a viewer control speed and direction through the corridors. Mohl called this the "first, embryonic version" of the system and listed its limits: the corridors lacked landmarks, the film quality was poor, and the hardware was primitive.

Nicholas Negroponte, the head of ArcMac, found support for a full movie map from DARPA's Cybernetics Technology Office, then headed by Craig Fields; Naimark describes the work as an academic, non-commercial and non-classified project. Mohl's thesis cites DARPA contract MDA-903-78-C-0039 and lists Negroponte as principal investigator and Lippman as project director. Lippman later recalled the goal as creating "so immersive and realistic a 'first visit' that newcomers would literally feel at home, or that they had been there before", and linked DARPA's interest to the Israeli rescue at Entebbe in 1976, for which soldiers had rehearsed on an airfield made up to look like the target.

Choice of Aspen and filming

Mohl gives the site criteria: a town of manageable size, geographically bounded and with distinct landmarks. Aspen covered about one square mile, had only three access routes, was bounded by rivers on three sides and a mountain on the fourth, and had a rectangular street grid aligned with the cardinal directions. As a resort, it also had a large pool of first-time visitors who could serve as test subjects. Naimark adds personal reasons: his sister lived in Aspen, Mohl had grown up in Colorado and photographed around Aspen, and the Aspen Design Conference was a possible venue for presenting the result.

The town's roughly 20 miles of streets were filmed three times: in fall 1978, in winter 1978-79 and in fall 1979. Naimark's account says the last pass was brief and used an active gyro stabilizer, that filming ran daily between 10 AM and 2 PM to limit lighting differences, and that the vehicle drove down the center of each street so that cuts would match. The camera rig was designed and operated by wildlife cinematographer John Borden, MIT undergraduate Stan Syzaki and Naimark.

Mohl's thesis records the capture methods in detail:

Capture element	Details (as described by Mohl, 1981)
Filming rig	Van with a camera platform 11 feet above the ground; a higher viewpoint was chosen because lower ones put distracting emphasis on pavement and vehicles
Frame trigger	A bicycle wheel trailing the rig triggered the cameras at fixed distances
Frame interval	One frame every 10 feet, chosen so the footage would fit within the 50,000 frames of one videodisc side; a 5-foot interval was also tested
Footage per pass	15,000 frames of straight sequences and 12,000 frames of turns
Cameras	Standard arrangement of three 16 mm Bolex EL cameras aimed along and perpendicular to the direction of travel; a fourth, rear-facing camera used on the first 1978 shoot was dropped as redundant
Stabilization	Three camera suspensions were tested: two counterweighted designs by John Borden of Peace River Films, and an inertially stabilized rate gyro with servo-controlled hydraulics from Tyler Camera Systems
Panoramas	An anamorphic Volpi lens on a 35 mm camera pointing straight up, recording a 360-degree panorama from 30 degrees above to 30 degrees below the horizon
Turns	"Pivot turns" shot every 10 degrees at an intersection (9 frames for a right-angle turn), "truck turns" shot as the vehicle turned (the standard technique eventually adopted), and hybrid "leading turns"
Facades	2,000 facades (including parks and other undeveloped sites) photographed on 35 mm in both fall and winter with matching registration

The street footage was only the backbone of a broader survey of the town. It included short documentaries by MIT film professor Richard Leacock, binaural sound recorded by Steve Gregory and Rebecca Allen, tens of thousands of still frames, historical images re-photographed by Scott Fisher, non-audiovisual statistics gathered by Walter Bender, and overview-map ideas from UC Santa Cruz cognitive psychologist Kristina Hooper. MIT Technology Review reported that the finished map let users listen to interviews with locals and "go back in time" to historic photos of existing buildings.

Completion and publication

The material was edited at MIT and mastered onto videodiscs, and the controlling software was written with the help of graduate students including Steve Yelick, Paul Heckbert and Ken Carson. The system was ready for its first demonstration in summer 1979. Four discs were mastered in all; the fourth combined street travel in different seasons, aerial overviews, facades, cultural material, computer animation, 360-degree panoramas and computer-synthesized turns. Lippman presented the work at SIGGRAPH 1980, and Mohl submitted his thesis in September 1981, receiving his Ph.D. in Education and Media Technology in February 1982.

Playback system

Mohl's thesis documents the hardware. A 32-bit Perkin-Elmer minicomputer, standard equipment at ArcMac, held a database of the town's topology (each block and turn linked to its start and end frame numbers on the disc), the aerial-overview frames, and the name, address and facade frame of every building. Mohl remarked that most of the project could have run on a small personal computer.

Video came from MCA DiscoVision PR-7820 players with a maximum seek time of 5 seconds. To hide that delay, the system used two players: while the viewer watched one, the other searched for the next sequence, and a video switcher cut between them. Mohl names this "two-disc algorithm" as one of the two most general techniques the project produced; with careful disc layout it made up to four branching choices available without any search, for example by placing a left turn and a right turn on the same starting frame, one sequenced forward on the disc and the other backward. Graphics from a 9-bit Ramtek frame buffer were keyed over the disc video to draw menus and navigation aids, and a Votrax speech synthesizer announced street and building names.

The user station had a color monitor with an Elographics transparent touch overlay. An earlier version used an x-y force joystick, and a body-worn Polhemus position sensor (ROPAMS) was also tried as an input device. Pressing further along the on-screen speed bar stepped the disc faster, up to 10 frames per second, which at 10 feet per frame corresponded to an apparent 70 mph.

Computer-generated model

Alongside the film, ArcMac built an abstract three-dimensional model of Aspen, the Quick and Dirty Animation System (QADAS). Building footprints were digitized from topographic maps with a tablet and heights estimated from the number of stories seen in the travel footage. The renderer had seven degrees of freedom (position, orientation and lens focal length), drew polygons back to front with the painter's algorithm, and took 30 seconds to one minute per frame; each frame was photographed onto 35 mm film for transfer to videodisc. Photographs of prominent facades were digitized, cleaned up and mapped onto the building faces to add realism, and the model was used to film viewpoints a real camera could not reach, such as vertical moves from ground level to 600 feet and "helicopter" approaches to the town. Lippman's 1980 abstract describes this choice between "sparsely sampled sequences of images" from single-frame cameras and "computer synthesized replicas of those images".

The team also generated "synthesized turns" by computer: two perpendicular views down intersecting streets were digitized and the in-between views were interpolated, with or without perspective correction, and similar turns were computed from the 360-degree anamorphic panoramas.

Spatial learning research

Mohl's thesis studied the "cognitive maps" that people formed of Aspen: residents, ordinary visitors, movie map users who had never been to Aspen, and movie map users who then visited Aspen for the first time. His abstract states that "movie map training leads to superior way-finding competence in the real setting", and that users who also had personalized reference maps showed significant differences and some advantages.

In a controlled experiment, 24 subjects drove the same 20-block route six times, either with routing diagrams overlaid on the aerial map beforehand ("map-plus-travel") or with street travel only. Landmark recognition rose from just above chance (60 percent) to 80 percent by the fifth trial. Views identical to those seen on the route were recognized with 88 percent accuracy, rotated views with 71 percent, and distractor images from a similar parallel street were rejected at only chance level (53 percent). The map-plus-travel group did better on landmark recognition and on metric distance estimates, while the street-only group showed better sequential route knowledge. Four subjects who used the movie map before visiting Aspen reported that it had prepared them well for specific views and spatial relations but poorly for scale: mountains were larger and closer, and some streets steeper, than they expected. One subject formed a mistaken picture of a T intersection beside a park during surrogate travel and searched for it on the wrong side of the park throughout her four-day stay; Mohl concluded that the medium's persuasiveness was "both a virtue and liability".

From these studies Mohl recommended that future playback spaces have a large field of view, a peripheral field built on curved surfaces, and "spatial correspondence" between the displayed image and the viewer's head and eye movements. He also wrote that "no one is fooled into thinking the video display is a window onto a live reality".

Later moviemaps

Naimark recalls that moviemapping looked at the time like a coming medium for virtual travel and tourism. Several members of the Aspen team later made moviemaps of Paris for the Paris Metro (1985), Palenque for the Bank Street College (1985), San Francisco for the Exploratorium (1987), Karlsruhe for the Center for Arts and Media (ZKM) (1990), and Banff for the Banff Centre for the Arts (1993). In the early 1990s the Apple Multimedia Lab organized a one-day "sweep" of Moss Landing, California, and in the mid-1990s the UC Berkeley Center for Design Visualization and the UNESCO World Heritage Centre led groups to sweep cultural heritage regions in Europe.

Influence and legacy

In a Computer History Museum history of surrogate travel and Google Street View, Marc Weber writes that the project "pioneered basically all of the features of street view and other mapping services today", including navigation buttons for turning and moving, integration with flat maps and aerial photography, computer-generated panoramas and 3D models of buildings, and describes how Google co-founder Larry Page became interested in capturing street-level views at scale in 2003. MIT Technology Review described the 1979 system as anticipating Street View "almost 30 years" earlier. An MIT Media Lab article by David Young, a former research assistant there, credits the project with turning an early laserdisc player from a linear movie player into a randomly accessed medium under user control.

Academic work has also examined the project as history. In Television & New Media, Aubrey Anable set aside its well-known military provenance and read it as an expression of ArcMac's ideas about the individual in human-computer interaction, formed in response to 1970s discourses of urban crisis and "techno-paranoia". Janina Hoth's 2021 paper at the ARTECH conference studied the collaboration between artists and engineers at ArcMac, describing the movie map as "a predecessor of Google Street view and other navigation technology" and one of the first tools in which users could navigate freely through a virtual map.

Researchers working on video-based virtual travel still cite the project directly. In 2020 Naoki Sugimoto, Yoshihito Ebine and Kiyoharu Aizawa built a new movie map from omnidirectional video, noting that the only earlier movie map prototype had been developed with analog video technology; in their user study the system gave a better user experience than Google Street View for finding a landmark. A later paper with Sugimoto and Aizawa among its authors, published in Multimedia Tools and Applications in 2026, calls the Aspen Movie Map "the seminal work of interactive videos" and builds avatar-based virtual worlds for urban exploration from 360-degree video.

The original material is also being reused. A laserdisc copy (ASPEN4, the software version Lippman used in later demonstrations) is preserved on the Internet Archive. An ongoing MIT Media Lab project by Yufeng Zhao, "Aspen Movie Splat", is reconstructing 1970s Aspen as a hierarchical 3D Gaussian splat scene from high-resolution scans of the original 16 mm reels. The laserdisc rip proved too noisy for feature matching, but structure-from-motion processing of the film scans aligned about 27,000 of roughly 36,000 driving frames.

See also

• History of virtual reality
• QuickTime VR
• 360 Video
• Google Earth VR
• Telepresence