SPACE FORTRESS: HISTORY
Space Fortress is a complex-skill task developed in the Cognitive Psychophysiology Laboratory at the University of Illinois for the Learning Strategies project. Its early development was iterative: computer-controlled, game-like tasks were refined through experiments before the version documented in 1989.
That canonical description was published by A. Mané and Emanuel Donchin in Acta Psychologica. It credits a development process within the laboratory rather than presenting the game as a single-author work. The task was made deliberately demanding: continuous steering, targeting, temporal control, and concurrent monitoring had to be coordinated over extended practice.
From laboratory task to game family
Across its history, Space Fortress has been adapted for different studies and platforms. Full task variants introduced friend and foe mines, finite missile resources, a stream of monitor symbols, bonus choices, and detailed component scores. Later work also used practice-oriented training structures and revised the task for contemporary computer environments.
The 2018 open-source reinforcement-learning implementation made a compact version available as a benchmark, including player-turn and auto-turn modes. That codebase is the behavioral reference for this browser prototype.
This web build concentrates on the central flight-and-fortress loop: inertial navigation through a hexagonal corridor, timed shots, fortress vulnerability, and a three-minute round. Mines, monitoring, resource management, audio, and other historical task layers are possible future game modes; the project README records those options and their source descriptions.
Primary history and reading
- Mané & Donchin (1989), “The Space Fortress Game” — original task description and its complex-skill acquisition purpose.
- Lee et al. (2010), cognitive-control training in older adults — an open-access description of a later, fuller task variant and its timing rule.
- Boot et al. (2016), individual skill structure — a study using archival Space Fortress performance data.
- Miller & Boot (2018), looking behind the score — open-access work on skill structure in the task.
Software and benchmark
- Agarwal, Hope & Sycara’s GPL-3.0 Space Fortress repository — the behavioral/configuration reference for this project.
- Challenges of Context and Time in Reinforcement Learning (2018) — introduces Space Fortress as a reinforcement-learning benchmark.
- Learning Time-Sensitive Strategies in Space Fortress (2018) — related open preprint on the timing challenge.
Links are provided for scholarship and provenance. Individual papers and code repositories retain their own licences and access conditions.