Immersion is how humans actually learn languages.

From the beginning, SonaXR's speech, scoring, and measurement pipeline was built to be language-agnostic. The English demonstration is the proof: the same architecture that scores European Portuguese now runs an English-language conversation scenario, with a Non-Player Character that responds to the participant's speech in real time within a scripted, phase-structured flow, and a measurement layer that evaluates the participant's spoken English as it happens. The dialogue is authored and phase-gated, not open-ended free conversation. English is not a separate product line — it is the existing platform pointed at a different target language.

Because the instrument is the same, the data is the same. Every English session produces the same objective performance record the platform generates for any language: phoneme-level pronunciation scoring, per-word accuracy, fluency measures, response timing, and the session transcript, written to a structured, audit-logged store and exportable into an organization's own analysis tools. The platform measures and exports; interpretation belongs to the deploying organization.

This matters because English proficiency is itself a research subject and a readiness factor across every audience SonaXR serves. The same English-evaluation capability is relevant to all of them — higher-education programs working with English language learners, research labs studying English speech production and acquisition in populations they recruit, and federal language programs for whom English proficiency is a measurable component of readiness. One instrument, one data format, applied wherever spoken English needs to be measured under standardized, reproducible conditions.

The English scenario is a demonstrated capability — a working build shown on hardware, not a released product — running on the same Meta Quest 3 deployment, the same character and conversation architecture, and the same privacy-by-design analytics — volatile session tokens never derived from a device identifier, raw audio stripped at the persistence layer before any write. What is true of the platform's measurement and privacy posture for any language is true here.

Immersion is the method the most intensive language programs are built on — the Foreign Service Institute and the Defense Language Institute among them. The civilian language acquisition market, by contrast, is dominated by consumer applications optimized for daily-streak engagement. SonaXR is building immersion in research-grade form — culturally authentic scenarios, contingent character behavior built against established presence and rapport constructs from Human-Computer Interaction research, and phoneme-level pronunciation scoring that produces per-student performance data classroom delivery does not generate. Instructor-facing reporting on top of that data is on the product roadmap.

SonaXR's planned higher-education model is institutional licensing for language departments, built on a study-enrollment and consent infrastructure designed to support cohort-based deployment, with Common European Framework of Reference-aligned skill measurement; instructor-facing progress reporting is on the product roadmap. Cultural authenticity is engineered into scenario design — the platform's first built scenario is a nine-phase European Portuguese kitchen breakfast dialogue with a native-register Non-Player Character, target-culture register and politeness norms, and contextual scaffolding that reflects how the language is used in the country it is spoken in. Students are not memorizing phrasebook fragments; they are practicing inside a place that responds to them. It is one scenario today; the scenario library is being built.

For multinational employers preparing employees for international assignments, corporate language training tied to measurable proficiency data is a planned offering. The architectural basis for it already exists: a consent and deletion-request tracking pipeline paired with session-token-only analytics.

Individual licensing is a planned tier, giving learners outside the consumer-app paradigm access to the same Virtual Reality conversation infrastructure. The product is positioned as a research-grade alternative to streak-based learning — oriented toward measured spoken performance rather than engagement streaks, with longitudinal session data exportable to the learner.

The primary development language is European Portuguese, used as a test bed during the build phase. The underlying speech, retention, and measurement pipeline is language-agnostic by design — demonstrated by a working English-language scenario that runs the same instrument against English speech. Additional target languages will be added as the pipeline is validated against each and partner contracts are in place.

Second-language learning is a widely studied candidate intervention for cognitive reserve, and the evidence is genuinely contested: meta-analyses report a delay in dementia onset among bilinguals, while prospective incidence data do not reach significance. We do not claim the protective effect is established. We claim the field lacks a standardized, measurable, reproducible way to deliver and instrument the intervention — and that this is one reason the question stays open. Most consumer tools are optimized for engagement metrics rather than Institutional Review Board-grade outcome measurement. SonaXR is being built to close that gap, in cognitive aging and in the other speech research areas where standardized spoken-interaction data is scarce.

The platform is engineered against the validated measurement instruments cognitive science and Human-Computer Interaction research already use. Place Illusion and Plausibility Illusion (Slater, 2009; Slater et al., 2022) define the construct of Virtual Reality presence the platform is designed to deliver. The Networked Minds Social Presence Inventory (Biocca and Harms, 2003) and the Avatar Embodiment Questionnaire (Peck and Gonzalez-Franco, 2021) are the standardized social-presence and embodiment instruments the platform is built to be measured against. The Tickle-Degnen and Rosenthal three-factor model of rapport as operationalized in virtual-agent research (Gratch et al., 2007; Huang, Morency and Gratch, 2011) informs the platform's character-behavior architecture. These constructs are well-established in their respective literatures. SonaXR itself has not yet been scored against any of these instruments — doing so, alongside validating the platform’s behavioral affect proxy against established affect measures, is a first-order goal of early pilots and an open collaboration opportunity for an incoming investigator.

Pilots are structured as Institutional Review Board-reviewed protocols, with study design — duration, cohort size, intervention dosage, and outcome measures — left to the principal investigator. SonaXR captures and exports objective performance data; the investigator defines every research question and performs every interpretation. The platform does not diagnose, screen, monitor, or treat any condition, and makes no clinical claim about any participant. Proposed pilot pricing is set to sit below the procurement thresholds typical at many institutions. The platform's privacy architecture generates cryptographic session tokens that are never derived from device identifiers, and raw audio is stripped at the persistence layer before any analytics record is written. A deletion-request tracking pipeline with a dual-authorization gate and audit logging records data-subject deletion requests for institutional review. Institutional Review Board protocols at each site layer consent capture and retention policy on top of this architecture.

SonaXR is initiating discovery conversations with speech, second-language acquisition, and cognitive aging researchers, including Alzheimer's Disease Research Centers affiliated with the National Institute on Aging's research network. A National Institutes of Health Small Business Innovation Research (SBIR) Direct-to-Phase-II application is in preparation for an upcoming receipt date. Direct-to-Phase-II does not require a prior Phase I award but does require demonstrating that feasibility has already been established; SonaXR is assembling that evidence base from its working prototype. No award has been made and no application has yet been submitted.

The Department of War invests heavily in language capability — through the Defense Language Institute Foreign Language Center, through the service academies, through career-long sustainment programs across the Air Force, Army, Navy, Marine Corps, and Coast Guard, and through Special Operations Command's regionally aligned forces. Across that span, two persistent challenges recur: delivering immersive, conversation-based instruction at scale, and measuring proficiency continuously rather than at end-of-cycle examinations.

SonaXR addresses both. The platform delivers immersive Virtual Reality conversation in target languages, with structured phases that move learners from elementary exchange through more complex discourse. Embedded measurement runs continuously and produces objective data: phoneme-level pronunciation scoring, vocabulary retention via an SM-2-family Spaced Repetition System, and Common European Framework of Reference-aligned progression, which maps to the Interagency Language Roundtable scale through published approximate crosswalks. Speaking is not part of the Defense Language Proficiency Test — the Oral Proficiency Interview is what evaluates spoken proficiency, and it is the part of the federal language assessment system the platform is most relevant to. The platform's spoken-practice session structure is designed around the same warm-up → level-check → probe → wind-down shape the Oral Proficiency Interview uses, with functional task ladders (describe → narrate → handle situation → support opinion) aligned to Interagency Language Roundtable level descriptors. The platform produces measurement data to inform a program's own view of readiness; it complements formal assessment rather than replacing it, and proficiency determinations remain with the assessing authority.

Use cases span the full lifecycle. Initial acquisition practice for entry-level linguists at the Defense Language Institute Foreign Language Center. Sustainment practice between formal training cycles — structured, measurable practice for personnel who must hold a required Interagency Language Roundtable level (commonly S-2 or S-3) when operational requirements demand it. Pre-deployment refresh for regionally aligned units. Long-tail maintenance for personnel who acquired a language years prior and want measurable evidence of continued practice. The same Virtual Reality pipeline serves all of them, producing measurement data in formats structured for a program's own analysis.

Rapport architecture is the platform's hypothesis for producing voluntary practice behavior rather than compliance behavior. The reasoning is that a system producing a checkbox gets logged into; a system producing a relationship gets returned to. SonaXR's character behavior is engineered against the three-factor rapport model of Tickle-Degnen and Rosenthal (1990), as operationalized for virtual agents by Jonathan Gratch's program at the Institute for Creative Technologies (Gratch et al., 2007; Huang, Morency and Gratch, 2011) — work funded by the Defense Advanced Research Projects Agency and the Army Research Laboratory. Whether that architecture actually raises voluntary practice rates in an operational population is an empirical claim SonaXR has not yet tested, and measuring it is a proposed objective of any pilot.

The underlying pipeline is language-agnostic. The current development language is European Portuguese, used as a test bed during the build phase and not yet fully implemented for production deployment. The same phoneme-level scoring and retention architecture is designed to support any target language — including Defense Language Institute Category III and IV languages — once the test-language implementation is fully validated and partner contracts define language priorities. Building on Unity XR is intended to support deployment across major enterprise headsets over time, though current builds are validated on Meta Quest 3.

SonaXR is preparing toward a Department of War Direct-to-Phase-II submission and is evaluating AFWERX Open Topic pathways. No Department of War contract, award, or pilot is currently in place. Direct-to-Phase-II requires demonstrating Phase-I-equivalent feasibility, and SonaXR is building the prototype evidence base accordingly. Engagement targets include the Defense Language Institute Foreign Language Center, the service academies' language departments, Air Force language programs, Army language programs, and United States Special Operations Command.

Most language software reports a streak or a score. SonaXR is built the other way around: measurement is the product. Every spoken interaction generates an objective performance record — phoneme-level pronunciation scores, per-word accuracy, fluency measures, response timing, and the session transcript — written to a structured store the moment it happens.

Because the task environment is standardized and reproducible, the same measures are produced the same way in every session. That makes the data comparable across sessions and across users, so a program can track performance on its own terms over time. SonaXR reports the measurements as objective numbers; what they are used for is decided entirely by the organization that deploys it.

Data leaves the platform through a structured, audit-logged export designed to drop cleanly into a program's own analysis tools. The export is keyed to a volatile cryptographic session token that is never derived from a device identifier, and raw audio is stripped at the persistence layer before any record is written — so the exported data carries performance measures without carrying identity. A deletion-request pipeline with a dual-authorization gate and audit logging is built in.

This measurement-and-export layer is the same across every SonaXR deployment — civilian, military, and academic alike. The platform measures and exports; it does not interpret. Any conclusion drawn from the data belongs to the organization or investigator that deploys the platform, under their own policies and protocols.