Comparing Varieties without Ranking Them | Catalan Grammar

A fair comparison holds context constant before interpreting a difference. If one recording is an edited Barcelona newscast and another is an informal conversation in an Alghero kitchen, almost every contrast confounds region with genre, age, audience, preparation, and contact history. The first methodological rule is simple: compare like with like.

Catalan is pluricentric and internally continuous. Central, Valencian, Balearic, Northwestern, Northern, and Alguerese varieties share a core while distributing phonological, morphological, lexical, and pragmatic features in overlapping bundles. A feature matrix can display those bundles; it cannot turn them into a ladder from “best” to “worst.”

Before comparing data, review the map of regional continua and the relationship between the general and regional standards.

Define the object before collecting examples

“Valencian” can mean the speech of a particular town, a broad dialect group, an institutional standard, or the language practices of everyone in an administrative territory. Those are not interchangeable datasets. Similarly, “young speakers” is not a place, and “formal Catalan” is not an accent.

State at least six variables:

locality or network;
speaker age and relevant language history;
interactional setting;
genre and degree of preparation;
date of recording or publication;
linguistic feature being measured.

Aquest estudi compara entrevistes informals enregistrades a Girona i a Castelló entre parlants de la mateixa franja d'edat.

This study compares informal interviews recorded in Girona and Castelló among speakers in the same age range. (academic register; a controlled claim)

⚠️ A València parlen més formalment que a Barcelona.

Unsupported comparison — it treats cities as uniform and does not control context or define formality.

The first sentence is narrower, but therefore stronger. It still needs sampling details; it does not claim to represent every speaker in either location.

Build a feature matrix, not a stereotype list

A useful matrix uses comparable structural questions and allows variation inside cells. It separates the codified regional standard from spontaneous local use.

Feature	Central baseline	Valencian	Balearic	Northwestern / Northern / Alguerese caution
unstressed a/e	usually merged in speech	generally distinguished	Eastern systems, with insular distributions	Northwestern generally Western; Northern and Alguerese require local analysis
1sg present of parlar	parlo	parle	parl	do not extrapolate one verb to a whole paradigm
demonstrative	aquest	este and aquest are codified; choice varies by area and model	aquest, with local phonetic forms	form, distance system, and register must be recorded separately
articles	general el/la	general el/la	salat system in many local/informal domains; general article in others	Northwestern lo and Alguerese systems need their own domains
weak pronouns	em, ens, us; productive en/hi	regional ordering and shapes within a coherent system	me/mos/vos occur regionally; distributions vary	measure actual combinations, not one citation form

The matrix intentionally contains cautious cells. False precision is not better than an explicitly delimited comparison. A claim about Alguerese based on one speaker or about Northern Catalan based on a school grammar should be labeled provisional.

💡

A matrix is an index to evidence. Each cell should point to tokens, speakers, registers, dates, and coding decisions; otherwise the table merely gives stereotypes a scientific appearance.

Matched texts isolate variables

Translation or elicitation can hold propositional content constant. It may not reproduce spontaneous frequency, so combine it with natural corpora.

Aquesta tarda començo a preparar la meva part del projecte, però encara no en tinc totes les dades.

This afternoon I begin preparing my part of the project, but I still don't have all the data for it. (Central-oriented standard)

Esta vesprada comence a preparar la meua part del projecte, però encara no en tinc totes les dades.

This afternoon I begin preparing my part of the project, but I still don't have all the data for it. (regional: Valencian standard)

Aquesta tarda començ a preparar sa meva part del projecte, però encara no en tenc totes ses dades.

This afternoon I begin preparing my part of the project, but I still don't have all the data for it. (regional: Balearic local/informal article model)

As an audio-ready task, align recordings and listen separately for vowels, first-person endings, articles, possessives, and rhythm. Do not ask vaguely which speaker has “more accent.” The matched sentences also show shared syntax and partitive en, preventing salient differences from hiding common structure.

Corpora answer frequency questions—with limits

A corpus can test whether a form occurs, where, and how often under its sampling design. It cannot by itself show that a rare form is unacceptable, or that a frequent social-media form belongs in formal law. Written corpora overrepresent edited genres; broadcast corpora may overrepresent trained speakers; online searches duplicate content and obscure demographics.

Al corpus periodístic, «meua» apareix sobretot en mitjans valencians; això descriu una distribució regional, no una manca de formalitat.

In the journalistic corpus, meua appears mainly in Valencian media; that describes a regional distribution, not a lack of formality. (academic interpretation)

La forma és poc freqüent en aquesta mostra, però el corpus no inclou conversa espontània de les Illes.

The form is infrequent in this sample, but the corpus does not include spontaneous conversation from the Balearic Islands. (sampling caveat)

Normalize counts, inspect concordance lines, remove duplicated texts, and report metadata gaps. For weak pronouns, search functions rather than spellings alone: clitics contract and combine, so a simple word count can miss relevant tokens.

💡

Absence from a corpus means “not found under this corpus and query design.” It does not mean “impossible in the language.” Check tokenization, clitic spelling, genre coverage, and speaker metadata first.

Control register before claiming geography

Compare news with news, peer conversation with peer conversation, and administrative writing with administrative writing. A public-register sentence can be regionally marked and fully formal:

La Generalitat comunicarà demà les mesures i en publicarà el calendari complet al portal institucional.

The regional government will announce the measures tomorrow and publish their full timetable on the institutional portal. (formal public register; region-neutral structure)

La Generalitat comunicarà demà les mesures i en publicarà el calendari complet al portal institucional valencià.

The regional government will announce the measures tomorrow and publish their full timetable on the Valencian institutional portal. (formal public register; geographic reference does not change grammar)

When a Valencian institution writes hui or seua, record those as formal regional choices. When a Central friend says bueno repeatedly, record an informal discourse feature. Never compare the first to the second and conclude that one region is more conservative or correct.

Mutual intelligibility is task-dependent

Intelligibility is not one percentage inherent in a dialect pair. Recognition changes with topic, noise, speed, prior exposure, literacy, and whether the task asks for gist or exact transcription. Asymmetry is possible when one community receives more media from another.

A strong study distinguishes:

word recognition from understanding the message;
familiarity from structural similarity;
attitude from actual performance;
receptive success from productive ability.

Entenc el sentit general de l'entrevista, però no identifico totes les formes pronominals.

I understand the general meaning of the interview, but I don't identify every pronoun form. (precise receptive self-assessment)

That speaker has partial, useful comprehension—not zero intelligibility. Repeated exposure can improve reception without causing the learner to adopt the variety in production.

Contact claims need a mechanism

Spanish, French, Occitan, Sardinian, Italian, and global English participate differently across Catalan-speaking communities. To call a feature contact-induced, establish a plausible source, chronology, social setting, structural correspondence, and diffusion pattern. Similarity is not sufficient because Romance relatives share inherited material and can innovate in parallel.

La coincidència amb el castellà orienta la recerca, però no prova l'origen del tret.

The coincidence with Spanish guides the research, but does not prove the feature's origin. (academic register)

Equally, contact origin does not determine modern value. An old borrowing may be fully standard; a recent switch may be deliberately expressive; a calque may remain restricted to informal speech. Etymology, grammar, and register are three separate annotations.

Ethical description and political labels

Community names and political identities matter to speakers, but they cannot substitute for linguistic evidence. Administrative borders may cut across continua, and the same label can group internally diverse speech. Quote speakers' self-identification respectfully, then describe structures with replicable criteria.

Avoid metaphors of purity, decay, contamination, and incompleteness. They convert social power into alleged grammar. Also avoid a softer hierarchy disguised as distance: “closest to the standard” is meaningless unless a particular codified model and set of weighted features are specified.

En aquesta mostra, el valencià formal manté «hui» i «seua», mentre comparteix la major part de la sintaxi amb el model central.

In this sample, formal Valencian retains hui and seua while sharing most syntax with the Central model. (descriptive, non-ranking comparison)

Receptive and productive goals

For receptive competence, build correspondence sets and listen to matched recordings. Learn to map parlo/parle/parl, meva/meua, avui/hui, general and salat articles, and clitic shapes onto shared functions. For productive competence, select a coherent standard grounded in a community; do not assemble one “pan-Catalan” sentence by choosing the most conspicuous form from each region.

A learner may later style-shift intentionally. That requires knowing the social distribution of each feature, not merely recognizing its spelling.

Practice: repair a flawed comparison

Flawed claim: “Balearic is the most informal Catalan because speakers say sa casa, while Valencian is heavily Spanish-influenced because it says este.”

Repair it in four moves:

define the article and demonstrative systems separately;
compare the same register in both territories;
investigate the historical distribution of each form;
remove value judgments unsupported by structure.

A defensible result: “Many Balearic varieties use the salat article in locally defined domains while also using the general article; Valencian este belongs to a historically grounded, codified demonstrative system. Neither form by itself measures informality or contact intensity.”

Common Mistakes

❌ Aquest locutor representa tot el valencià perquè treballa en una televisió valenciana.

Incorrect — one trained speaker represents one repertoire, role, and editorial model, not every Valencian variety.

✅ Analitzarem diversos locutors, gèneres i localitats, i delimitarem què pot representar la mostra.

We will analyze several speakers, genres, and localities and delimit what the sample can represent.

❌ «Parl» és menys complet que «parlo» perquè no té vocal final.

Incorrect — Balearic parl is a complete first-person form in its paradigm.

✅ «Parl», «parle» i «parlo» expressen la primera persona dins paradigmes regionals coherents.

Parl, parle, and parlo express first person within coherent regional paradigms.

❌ Un resultat de cerca amb més exemples demostra quina forma és millor.

Incorrect — search counts reflect indexing, duplication, genre, population, and platform behavior.

✅ Cal normalitzar les dades, revisar el context i descriure la mostra abans d'interpretar freqüències.

Data must be normalized, context checked, and the sample described before frequencies are interpreted.

❌ Les fronteres polítiques determinen automàticament les fronteres dialectals.

Incorrect — linguistic continua may cross borders, while mobility and institutions create new patterns.

✅ Farem servir els límits administratius com a dada social, no com a prova lingüística suficient.

We will use administrative boundaries as social data, not as sufficient linguistic proof.

Key Takeaways

Compare matched speakers, tasks, dates, and registers before attributing a difference to region.
Feature matrices summarize evidence; they do not rank communities or replace corpus tokens.
Corpora answer questions only within their sampling and metadata limits.
Mutual intelligibility depends on task, exposure, speed, and familiarity as well as structural distance.
Contact explanations require chronology and mechanism; resemblance and political labels are insufficient.
Build broad receptive mappings and one coherent productive model rather than an accidental mixture.

Now practice Catalan

Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.

Start learning Catalan →