A Questionnaire That Survives the Uzbek Field

A good questionnaire isn't a list of questions — it's a route the respondent travels easily and answers honestly. In Uzbekistan that route has a second layer textbooks skip: the same instrument must work in Russian and in Uzbek, in a Fergana Valley mahalla — the neighborhood community — and in a Tashkent apartment, for an older woman who reads Cyrillic fluently and for a young respondent reading Latin script. A questionnaire that looks flawless on paper falls apart in the field exactly at these seams. Below are first the seven core design principles, then the part that really decides your data quality: RU↔UZ translation and cultural sensitivity, without which even a perfectly worded question returns biased answers.

The seven principles every questionnaire rests on

These rules are universal, but in the Uzbek field each one is tested harder — because you collect data in two or three languages and across very different social contexts.

One question, one idea

Avoid "double-barreled" questions like "How satisfied are you with the price and the quality?" A respondent might be happy with the price but not the quality — and can't answer honestly. Split such wording into two separate questions. Double-barreled items are especially dangerous in translation: what reads as one idea in Russian often splits into two in Uzbek, and the translator quietly fuses them their own way.

Plain language, not bureaucratese

Write the way your audience speaks. Jargon, acronyms, and bureaucratic phrasing raise the share of "don't know" answers. In Uzbekistan this rule is doubled: the wording must be simple in both Russian and Uzbek. Bureaucratic Russian translated word-for-word becomes a heavy Uzbek construction that the interviewer ends up paraphrasing on the doorstep — and there goes your standardization.

Don't hint at the "right" answer

Wording like "Do you agree our service has improved?" nudges people toward yes. Keep it neutral: "How has the quality of service changed over the past year?" In our context, leading questions are doubly treacherous — more on why in the courtesy-bias section.

Question order

Start with simple, neutral questions and leave sensitive ones (income, religiosity, political views) closer to the end, once rapport is established. Early questions set the tone and shape willingness to continue. In a mahalla, where the interviewer is essentially a guest, the opening questions are also part of building trust.

Balanced scales

A rating scale should be symmetric: an equal number of positive and negative options plus a neutral midpoint. An unbalanced scale skews results on its own — and in Uzbekistan a cultural tilt toward agreement (below) is added on top, so symmetry and neutral verbal labels matter twice as much.

Skip logic

Don't make respondents sit through questions that don't apply to them. Conditional branching shows the interviewer only the relevant blocks — cutting time and reducing error. On paper, manual skips are a frequent source of bad data; in a digital questionnaire the logic is built in and fires automatically.

Always run a pilot

Before the field stage, test the questionnaire on 10–20 people from your target audience. A pilot exposes ambiguous wording, logic errors, and overly long blocks before they spoil the whole dataset. And — this is critical for Uzbekistan — the Uzbek version must be piloted separately, never assumed to be "clear enough because it's translated."

Rule of thumb: if a question hasn't survived back-translation and a pilot in Uzbek, it isn't field-ready — no matter how flawless it looks in Russian.

RU↔UZ translation is methodology, not "hand it to a translator"

The most common and most expensive mistake in the Uzbek field: the questionnaire is written in Russian, handed to a "language-capable" staffer or freelancer, an Uzbek text comes back — and off it goes to the field. That text almost always drifts in meaning. Translating an instrument is not a linguistic service; it is part of the methodology and deserves the same seriousness as the sample.

Why does meaning drift? Russian and Uzbek encode modality, politeness, and degree differently. "Somewhat agree" versus "completely agree" requires precise word choice in Uzbek, or the scale steps stop being equal. Impersonal Russian constructions acquire a person and a tone in Uzbek. Terms like "household," "income," and "service" have several Uzbek equivalents with different shades — and the translator picks one silently.

What drift looks like in practice

Take a question that looks harmless: "How satisfied are you with the work of the local authorities?" "Local authorities" can be rendered as the district hokimiyat, as the mahalla, or as "the bosses in general" — three different objects of evaluation. Right there the dataset splits: some respondents rate the mahalla chairman, others the district hokim. Or the word "income": the everyday term daromad means only wages to many people, excluding remittances from migrants and household-plot produce — which in Uzbekistan are often half the family budget. If the translator didn't know the intent, they aren't to blame; the procedure with no reconciliation step is. These are exactly the quiet divergences back-translation catches before they become systematic error.

Professional translation plus independent back-translation

The working standard is a four-step procedure, the same logic international programmes use (UNICEF MICS, World Bank studies):

1. Forward translation. An experienced translator renders the questionnaire from Russian into Uzbek, aiming at the respondent's meaning and register rather than literal words.
2. Back-translation. A different specialist, who has not seen the original, translates the Uzbek version back into Russian.
3. Reconciliation. A methodologist puts the source Russian and the back-translation side by side. Every divergence is a signal that the Uzbek wording strayed from intent somewhere. The team discusses and fixes it.
4. Separate pilot. The Uzbek version is tested on Uzbek-speaking respondents on its own — ideally with cognitive interviewing ("what did you understand this question to mean?").

It looks expensive until you compare the cost with redoing the field stage because half the dataset was collected on a different question.

Latin and Cyrillic are two audiences, not two fonts

Uzbek is officially in Latin script, and instruments are usually best prepared in it. But the text the respondent reads themselves — showcards, option lists, self-administered blocks — is a different matter. Older respondents and many in rural areas read Cyrillic, the script they learned, more confidently. If a showcard with answer options is set in Latin only, some people will "read" it from memory and guess.

Practical takeaway: keep the interview instrument in Latin, but provide showcards and self-administered text in both scripts, or match the script to the audience of the study. Field teams must read both fluently regardless.

One questionnaire is not one language

If you genuinely want valid nationwide data, the instrument must exist in every language you will field, and the interviewer's language must match the respondent's:

• Uzbek and Russian — the base pair for almost any national study.
• Tajik — for Samarkand and Bukhara, where a meaningful share of respondents are more comfortable in Tajik.
• Karakalpak — for Karakalpakstan this is not optional: Karakalpak is a distinct Turkic language, and without its own version the data from there will be incomplete and biased.

Each additional version runs the same cycle: translation, back-translation, reconciliation, separate pilot.

Scales and idioms are localized, not translated

Agreement scales and rating gradations are a special headache. The continuum from "completely disagree" to "completely agree" must land on words a native speaker actually uses, with equal semantic steps — not on a calque. A pilot with cognitive interviewing shows whether the respondent truly distinguishes "qisman roziman" (somewhat agree) and "toʻliq roziman" (fully agree) as adjacent rungs, or treats them as the same thing. The same goes for idioms and metaphors: a question like "how closely do you keep your finger on the pulse of…" becomes nonsense translated literally. Replace any figurative phrasing with a direct one, or find a local equivalent and test it separately. Numeric and "faced" (smiley) scales travel more easily than verbal ones — an argument in their favor for audiences of mixed literacy.

A multilingual team in the field: the instrument alone isn't enough

Even a perfectly translated questionnaire breaks if the interviewer's language doesn't match the respondent's. In practice this is a staffing and routing question: in Samarkand and Bukhara keep interviewers fluent in Tajik; in Karakalpakstan, native Karakalpak speakers; in the Fergana Valley most interviews will run in Uzbek, in Tashkent a notable share in Russian. Decide in advance who enters open-ended answers in which language, and how you will later code answers that arrive in three languages into a single scheme. If an interviewer translates a question off an Uzbek showcard "into Tajik from memory," you lose standardization again — better to have a ready version than to hope for improvisation.

Cultural sensitivity: the politeness that distorts your data

Uzbekistan has a strong culture of hospitality (mehmondoʻstlik) and respect for the person asking questions. For rapport this is a gift — the respondent is glad to receive you, offer tea, and help. But it carries a methodological cost: acquiescence, or courtesy, bias. Many people agree with a statement or give the "convenient" answer simply out of politeness, not to upset a guest or seem impolite. That means neutral wording matters not less but more here than in Western fieldwork.

Designing against courtesy agreement

• Avoid statements in the "Do you agree that…" format — reframe as a neutral question with symmetric options.
• Balance your scales and occasionally flip the polarity of statements in a battery, so an automatic "yes-yes-yes" becomes visible in analysis.
• State in the preamble that "there are no right answers here" and that the person's own opinion is what matters.
• Train interviewers not to nod or lead with intonation — especially in a warm, hospitable setting where slipping into a friendly hint is easy.

Sensitive topics: income, religion, politics, family

Direct questions about income, religiosity, political views, and intra-family matters draw many refusals and "socially desirable" answers in our context. A few working techniques:

• Ask about income in bands ("which range do your household's monthly expenses fall into?") rather than an exact figure — it's both more accurate and less anxiety-inducing.
• Frame religion and politics as neutrally and indirectly as possible; in some regions (Karakalpakstan's politically sensitive context is the clearest example) certain topics are wiser to leave out entirely than to collect unreliable or risky answers.
• Family topics demand tact and privacy — covered below, in the gender section.

Respect and register

Uzbek is rich in politeness forms. A young interviewer addressing an older respondent must use the respectful register — it affects both willingness to participate and answer quality. Build this into the wording and the interviewer instructions, rather than leaving it to improvisation.

The preamble and consent are part of the questionnaire too

Designing the introduction matters more than it looks for the Uzbek field. In a mahalla a stranger with questions is met with caution, so the preamble must briefly and honestly explain who is running the study, why, how long it takes, that answers are confidential, and that declining carries no consequence. Mentioning an official support letter (from a hokimiyat, a ministry, or a recognized client) and that the visit was cleared with the mahalla chairman often helps here — without it, some respondents simply won't start the conversation. But rapport cuts both ways: the warmer the welcome, the stronger the courtesy bias, so state plainly in the preamble that you need an honest personal opinion, not a "convenient" answer.

Consent must be delivered in the respondent's language — not a formality but a requirement of research ethics and Uzbekistan's personal-data law. If a person is more comfortable in Uzbek or Tajik, the consent text and the explanation of purpose must be in that language, or the "informed" part of informed consent is fiction. How to collect and store consent and sensitive data lawfully we cover in the piece on personal-data protection.

Showcards at the Latin–Cyrillic seam

Showcards deserve separate attention because the respondent, not the interviewer, reads them. A few practical rules for the Uzbek field:

• Keep options on a card short and parallel in structure — long phrasing in an unfamiliar script gets finished by guesswork.
• For an audience that includes older and rural respondents, keep a Cyrillic version of the cards ready; young and urban respondents read Latin confidently.
• Numeric and "faced" scales on a card remove part of the script problem — they're easier to "read" than verbal gradations.
• Test legibility in the pilot specifically with people for whom the script is unfamiliar, not with office colleagues.

Gender: reaching women without breaking the norm

In conservative and rural households a male stranger may not be received, and a woman may not answer, or may answer guardedly, in the presence of the male head of household. Add labour migration: a large share of working-age men (especially from the Fergana Valley) are working in Russia and Kazakhstan, so "who's home" tilts toward women, the elderly, and the young.

Practical responses:

• Female interviewers are often indispensable for properly surveying women — especially on health, family, and other sensitive topics.
• Plan the mode and place of the interview so a woman can answer without pressure from those present.
• Respectful register toward elders — distinctly for each gender and age.
• Spell out the respondent-selection rule within the household (Kish grid, last birthday), or the interviewer will default to whoever is convenient and home — and the sample drifts.

Assemble it into an instrument, not a stack of files

When you have two to four language versions, showcards in two scripts, skip logic, and selection rules, keeping it all in Word files is a path to field errors. A digital questionnaire ties the versions together: the interviewer picks a language, the logic fires itself, sensitive blocks can be hidden by condition, and the wording is identical across the team. You can assemble a multilingual instrument like this in the AISurvey builder — and test the branching before you pilot.

Next, related reading worth keeping beside this article: open vs. closed questions (how formats survive translation), sampling methods in Uzbekistan (who you'll even reach), and personal-data protection under Uzbek law (how to ask sensitive things lawfully). If you're just starting out, see the overview of AISurvey.

A short pre-field checklist

• Each question is one idea, neutrally worded, in plain language in both languages.
• Scales are symmetric, labels neutral, batteries include polarity flips.
• It passed the cycle: forward translation → back-translation → reconciliation → separate pilot of the Uzbek version.
• You have every language version you need (UZ, RU, plus Tajik and Karakalpak where relevant), with interviewer language matched to respondent language.
• Showcards and self-administered text account for Cyrillic readers.
• Sensitive topics are framed indirectly or dropped where sensible.
• Within-household respondent selection and the approach to women and elders are written down.

A questionnaire that passes this list really does survive the Uzbek field — and returns data you can trust.