Human Factors Issues in the Neural Signals Direct Brain-Computer Interface
Melody M. Moore, Philip R. Kennedy · 2000 · Proceedings of the Fourth International ACM Conference on Assistive Technologies (Assets '00) · doi:10.1145/354324.354351
Summary
This paper presents the software and human factors aspects of the Neural Signals direct brain-computer interface project — one of the first implanted BCI systems tested in humans. Unlike non-invasive BCIs that read EEG signals through the scalp, this system uses a neurotrophic electrode surgically implanted directly into the motor cortex. The electrode is a tiny hollow glass cone coated with a neurotrophic factor that encourages brain cells to grow into and through it, creating a stable long-term connection. Two gold wires transmit neural signals to a small amplifier and FM transmitter on the scalp, powered externally via induction coil with no batteries or wires penetrating the skin. By the time of this paper, three patients had been implanted: the first (MH) in 1996 demonstrated binary brain signal control before her death 76 days later; the second (JR, implanted spring 1998) achieved cursor control on a computer screen and could communicate using a virtual keyboard at approximately 3 letters per minute; the third (TT, implanted summer 1999) was still in early training. The paper describes the software ecosystem developed to support this work, including data acquisition tools, the Parmouse device driver that translates neural pulses into cursor movement, TalkAssist communication software, and a virtual piano used for training.
Key findings
The most remarkable finding was the emergence of what the researchers termed "cursor cortex" — patient JR initially controlled the cursor by imagining left hand movements, but over time the neural control migrated to other body parts (neck, eyebrow) and eventually became dissociated from any specific movement imagination. JR could not discern that he was moving any body part to drive the cursor, and when asked what he was thinking to control it, he spelled "NOTHING." This suggests the brain developed a dedicated neural region specifically for cursor control, demonstrating significant neuroplasticity. The researchers also observed "directionality" — JR appeared able to influence cursor direction (horizontal vs. vertical) from a single electrode, with different waveshape patterns and phase relationships between spikes corresponding to different movement directions. For communication, the team explored multiple approaches: a virtual keyboard where JR could spell at 3 letters per minute, iconic phrase-based communication (TalkAssist), and early experiments with Morse Code and phoneme-based speech synthesis. Training was found to be critical, with performance improving over time but degrading significantly with fatigue.
Relevance
This paper documented a pivotal moment in assistive technology — the transition from theoretical brain-computer interfaces to practical systems implanted in human patients with locked-in syndrome. The approximately 500,000 people worldwide with locked-in syndrome are conscious and intelligent but completely unable to move or speak, making direct neural interfaces potentially their only path to communication and computer access. The discovery of cursor cortex had profound implications for BCI design: rather than requiring users to maintain specific motor imagery indefinitely, the brain may naturally develop more efficient neural control strategies over time. The research agenda outlined — signal understanding, patient training, navigation paradigms, neural prosthetics control, and communication — essentially defined the field of implanted BCI research for the following decade. The paper also raised important questions about interface design for neural control, noting that paradigms designed for mouse or keyboard input do not necessarily work well when the input signal has fundamentally different characteristics such as variable firing rates and limited directional control.
Tags: brain-computer interface · locked-in syndrome · neural prosthetics · assistive technology · human-computer interaction · motor cortex · neuroplasticity · alternative input