The Tactile Grammars Project: Formalizing Deafblind Sign Language Structures

The Tactile Grammars Project envisions a deep linguistic documentation and formal analysis of tactile sign languages used by deafblind individuals, focusing particularly on Protactile (used in the United States) and one comparative language (e.g., Swedish Tactile Sign Language). Unlike signed languages that rely on visual perception, tactile languages are produced and perceived through touch, movement, and spatial awareness on the body. This project would involve ethnographic fieldwork, video recordings (with multiple camera angles to capture hand and body positions), haptic sensor data, and close collaboration with deafblind community members who are native signers. The core of the project is to build a structured grammar—from phonology (e.g., contact types, handshape, movement paths) to syntax (e.g., how signs combine in space and time)—and then encode that grammar into a computational model that can parse tactile signing sequences. The resulting corpus, grammar handbook, and parser would empower educators, interpreters, and technologists working with deafblind individuals.

The deafblind community is one of the most underserved in linguistic research. Current sign language grammars are based on visual modality. Tactile signing has distinct phonological, morphological, and syntactic features that are unaccounted for. There is no comprehensive reference grammar for any tactile sign language, leading to barriers in education, interpreter training, and technology development. Deafblind signers often face miscommunication in interpreted settings because interpreters lack formal training in tactile grammar. Furthermore, the rise of haptic feedback devices and wearable communication tools could greatly benefit from a formal understanding of tactile language structure, yet no such model exists.

• Document at least 500 lexical signs and 200 grammatical constructions from Protactile (a tactile sign language used in the US) and one other tactile language (e.g., Swedish Tactile Sign Language).
• Develop a formal phonological model that accounts for tactile parameters (e.g., contact type, movement path, handshape with orientation).
• Create a publicly available digital corpus with annotations (ELAN) and metadata.
• Design a computational grammar using Feature-Based Context-Free Grammar for parsing tactile signing sequences.
• Produce a grammar handbook for educators and interpreters, including annotated examples and teaching materials.

• Not attempting to standardize or replace natural tactile languages.
• Not developing a full communication device (that would be engineering).
• Not covering all tactile languages worldwide; focus on North American Protactile and one other.
• Not aiming to create a translation system; only analysis and modeling.

• Software: ELAN for annotation, Python for computational modeling, NLTK/Spacy for parsing, Praat for acoustic analysis (if applicable).
• Hardware: High-definition cameras (multiple angles) for video recording; optional haptic gloves with motion sensors for fine-grained hand tracking.
• Materials: Consent forms (in accessible formats), field notebooks, transcription protocols.
• Methods: Participant observation, elicited narratives, structured interviews, phonetic transcription of touch (using a modified IPA for tactile signs).

The project is structured in three phases:

1. Fieldwork and documentation: Establish long-term collaborations with deafblind organizations. Record 20+ hours of naturalistic signing in various contexts (narratives, conversations, instructions). Annotate lexical and grammatical features using ELAN with tier-based coding for tactile parameters.
2. Linguistic analysis: Identify patterns in tactile phonology (e.g., how signs combine, role of contact location) and morphosyntax (e.g., verb agreement via body location, use of non-manual markers like finger pressure). Develop a formal grammar using Head-Driven Phrase Structure Grammar (HPSG) adapted for the tactile modality.
3. Computational implementation: Encode the grammar in a Feature-Based Context-Free Grammar (FBCFG) and implement a parser in Python. Test on held-out data, refine, and release the parser along with the corpus and grammar documentation.

• Access and trust: Deafblind communities may be wary of researchers. Requires long-term commitment and collaboration with deafblind-led organizations.
• Data quality: Tactile signing is co-produced; cameras may miss fine hand movements. Need multiple angles and haptic sensors.
• Small sample size: Limited number of fluent signers. Generalizability may be weak.
• Interpreter bias: If using interpreters, they may filter linguistic forms. Ideally, the researcher should be fluent in tactile signing or work with deafblind co-researchers.

• How do tactile signs handle simultaneous vs sequential structures compared to visual sign?
• Is there a universal tactile grammar across different cultures, or are they as diverse as spoken languages?
• Can computational parsing capture the role of shared context (e.g., the interlocutor's hand position as a grammatical marker)?

The project required dedicated funding for a multi-year ethnographic study and close collaboration with a small, geographically dispersed community. Despite initial interest from a deafblind organization, the lead researcher moved to a different institution and the proposal was never resubmitted after losing momentum.

Idea originally conceived with Dr. X (a deafblind linguist) in 2018. The computational part was inspired by work on grammars of American Sign Language but adapted for tactile modality. Early pilot data from two signers suggested rich grammatical structures not found in visual sign languages.