L-MAS: the LINGACON Multilingualism Assessment Standard
How many languages does a person speak?
A framework for measuring multilingualism at population level in a comparable way
More than 7,000 languages are spoken today, and much is known about which of them are endangered, how they are structured and where they came from. Comparable data on a simpler question, how many languages a person speaks, are much harder to find.
Surveys, censuses and databases do collect language data, but they use different questions, thresholds and definitions. A person counted as bilingual in one survey may not qualify in another, so results from different countries cannot be compared directly. For entire regions, among them the Arab states, Latin America and most of Asia, we found no standardized, comparable repertoire data. L-MAS is a proposal for closing this gap.
The LINGACON Multilingualism Assessment Standard (L-MAS) is a five-level framework for measuring multilingualism at population level. It ranges from a single survey question (Level 1, minimal marginal cost) through a national audit with tested proficiency (Level 4) to a longitudinal tracker of change over time (Level 5). Every level produces the same core metric, ALPS (Average Language a Person Speaks), so that results can be compared across countries, survey programs and time.
L-MAS is freely available. The Level 1 question can be added to an existing survey.
Each level builds on the previous one. Organizations choose the level that matches their resources and objectives.
| Level | Name | Items | Primary Output | Cost |
|---|---|---|---|---|
| Level 1 | The One Question | 1 question | National ALPS estimate | Minimal |
| Level 2 | The Module | 4 questions | ALPS + proficiency depth | Low |
| Level 3 | The Full Battery | 8 questions | ALPS + connectivity data | Moderate |
| Level 4 | National Audit | 12 questions, standalone survey | Calibrated ALPS + confidence interval | Variable |
| Level 5 | Longitudinal Tracker | Level 3 repeated | ΔALPS over time | Sustained |
“How many languages do you speak well enough to have a basic conversation? Please include your mother tongue and any other languages you can converse in, even if you don’t use them every day.”
Follow-up: “Could you name them for me?”, recorded as open text. If the count and the list disagree, the list takes precedence. The threshold is intended to correspond to level A2 of the Common European Framework of Reference for Languages (CEFR). The full wording of every level is given in the L-MAS paper.
L-MAS is part of LINGACON, an independent research project whose overarching theme is the study of human cooperation. The standard grew out of the work on a global estimate of ALPS: assembling that estimate showed how differently existing sources measure the same thing. The three preprints below are released under CC BY 4.0.
The population-weighted mean repertoire across 237 countries and territories, with 934 languages indexed: 1.57 languages per person (1.568), within a grade band of 1.42–1.70.
Hovhannisyan, A. (2026a). The ALPS Index: A global estimate of the average number of languages a person speaks. SocArXiv. https://doi.org/10.31235/osf.io/3mtsx_v1
Version 2 was submitted on 12 October 2026 and is awaiting moderation. It corrects an error in version 1 in the reading of the Afrobarometer Round 9 and Round 10 items on other spoken languages.
Hovhannisyan, A. (2026b). The LINGACON Multilingualism Assessment Standard (L-MAS): Definitions, thresholds, and a review of survey practice. SocArXiv. Preprint page (latest version): https://osf.io/preprints/socarxiv/n96g3. Version 1: https://doi.org/10.31235/osf.io/n96g3_v1
The probability that a listener drawn at random from the world’s 8.2 billion people understands a speaker drawn at random: 11.3%, and 7.5% when the two live in different countries. English carries 39% of this connectivity.
Hovhannisyan, A. (2026c). GLCI (Global Linguistic Connectivity Index): How connected is the world? SocArXiv. https://doi.org/10.31235/osf.io/g3vxj_v1
Supplementary materials, row-level data, the pair table and the analysis code for the ALPS and GLCI papers, deposited on the Open Science Framework. Licences are stated for each component.
Multilingualism and linguistic connectivity, measured to a common standard
L-MAS assessments can be carried out for organizations, cities and countries. They report two metrics: ALPS (Average Language a Person Speaks), the average number of languages people speak, and GLCI (Global Linguistic Connectivity Index), the probability that a person drawn at random understands another person drawn at random.
For organizations working across three or more countries. Measure the multilingualism of the workforce and its internal linguistic connectivity, and see where communication gaps and language-training needs lie.
For municipalities, states, provinces and metropolitan areas. Describe the linguistic landscape of the population as a basis for integration services, education policy and public communication.
For governments and national institutions. An L-MAS Level 3 or Level 4 assessment, producing publication-quality ALPS data with confidence intervals and a regional breakdown.
An example of an L-MAS assessment report for a fictional company.
Select your profile below. We'll ask the right questions and respond with a detailed proposal.
It is natural to assume that larger countries are harder and more expensive to survey. For multilingualism this is mostly not the case. The sample a survey needs depends little on the size of the population and much more on how varied people’s language repertoires are.
Japan has about 124 million people, and most of them speak one language (ALPS 1.20). A standard national sample of 2,000–3,000 respondents can represent the population.
Nigeria has about 233 million people and more than 500 languages. Hausa, Yoruba and Igbo are the largest first languages, English is the official language, and Nigerian Pidgin is widely spoken as a second language. On average a person in Nigeria speaks 2.30 languages (ALPS, based on Afrobarometer Round 4 survey data), and which languages they speak depends largely on region and community. A representative sample has to be stratified by region and language community, the questionnaire has to be offered in several languages, interviewer training is more demanding, and data cleaning takes longer.
Nigeria has about twice Japan’s population, but its larger and more expensive survey design follows from the variety of its language repertoires much more than from its size.
The same holds at other scales. A linguistically diverse city such as Toronto, Dubai or Singapore is harder to assess than a homogeneous one, and an organization working in 15 countries needs a more elaborate design than one working in 3.
This relationship matters for budgeting, planning and interpreting L-MAS assessments, and cost estimates for an assessment take it into account.
For research inquiries, partnership proposals, media requests, or any other questions.