Survey Fatigue in Grid Questions: What Straightlining Costs

Survey fatigue shows up first in grid questions, as straightlining. What the research measures, why cleaning cannot recover the data, and how to change forma...

The brand block starts deep into the questionnaire. Brand names down the side, an agreement scale across the top, and a battery of statements to rate one brand at a time. It is the block the client cares about most, and it sits after the usage questions, the purchase questions and the media questions. Survey fatigue is what happens to the respondent by then: the answers keep arriving, and the attention behind them has gone. In a grid it has a recognizable shape. The same column, all the way down.

Key Takeaways

• In a web survey experiment that randomized the order of question blocks, Galesic and Bosnjak found that questions placed later in the questionnaire got faster, shorter and more uniform answers (Public Opinion Quarterly, 2009).
• In the LifeOnSoMe study of 3,285 adolescents answering 64 items, 5.4 percent were straightliners, and among them the internal consistency of one scale reached 0.99 against 0.79 in the rest of the sample.
• Across 14,613 web responses to a health registry survey, respondents were significantly more likely to straightline in all five traditional grids than in a stacked format showing one item at a time.
• Cleaning after fieldwork can flag or drop a straightliner. It cannot recover the answers that respondent never gave, so the fix belongs in the questionnaire design.

What survey fatigue looks like in the data

Survey fatigue is the drop in attention and effort as a respondent moves through a long questionnaire. It rarely shows up as an empty answer. It shows up as answers given faster and with less thought.

A grid question, also called a matrix question, is a block of items that share one response scale, with the items as rows and the scale as columns, all on one screen. Straightlining is giving the same answer to every item in a series, or, in the partial form, to an unusually long run of consecutive items. A speeder is a respondent whose completion time is so far below the median that the questions could not have been read.

The clearest evidence that position alone causes this comes from an experiment by Mirta Galesic and Michael Bosnjak. They stated different lengths for the same web questionnaire, 10, 20 or 30 minutes, and presented blocks of related questions in random order. The longer the stated length, the fewer people started and finished. And the same questions, placed later, got answers that were faster, shorter in the open ends and more uniform in the grids.

Because the order was random, the content of the questions cannot explain the difference. The only thing that changed was how much questionnaire came before them.

How the long questionnaire is built today

PERSONAL EXPERIENCE In most of the quantitative studies I have seen, the long questionnaire is still a sequence of grids. The brief comes in, the questionnaire is written in a document, and the programmer turns each battery into the grid template the collection platform offers. Every study gets the same screen: a table of radio buttons and a progress bar at the top. A packaging test looks exactly like an employee survey.

There are reasons. Grids save space and keep the questionnaire short on paper. In trackers they keep the instrument identical to previous waves. The Survey Practice paper on smartphone grids gives both reasons for keeping the grids in a health registry's fourth wave: comparability across waves and across web and mail modes.

The cost of that choice shows up only in the data file. The completes arrive, the quota fills, and nothing flags the respondent who answered "agree" to every statement for every brand. That respondent counts as a complete.

What the research measures about grids and straightlining

The LifeOnSoMe study, published in Scientific Reports in 2025, followed 3,285 adolescents aged 16 and over in Bergen, Norway. They answered 64 consecutive items about social media use, all with the same five response options. A straightliner was anyone whose longest run of identical consecutive answers was a statistical outlier, which in that sample meant 16 or more in a row. That was 5.4 percent of participants, 8.6 percent of boys against 2.9 percent of girls.

The authors then measured what those respondents do to a scale. Cronbach's alpha is the standard measure of how consistently the items of a scale move together, from 0 to 1, read as reliability. Straightlining inflates it, because identical answers are perfectly consistent.

Source: Skogen and colleagues, Straightlining prevalence across domains of social media use, Scientific Reports, 2025, Table 2.

Read the last row. The scale that looked weakest among attentive respondents looked strong among the ones who stopped reading. In that sample the straightliners were few and the overall conclusions barely moved, which the authors say plainly. They also warn that the discrepancies would matter in samples where straightlining is more common, and long commercial questionnaires are where it gets common.

Straight lines make a scale look more reliable

Paired horizontal bars of Cronbach’s alpha, on a scale from 0 to 1, for six social media domains, comparing the sample without straightliners with straightliners only. Self-presentation: 0.87 without, 0.95 among straightliners. Negative acts and exclusion: 0.79 without, 0.99 among straightliners. Unwanted attention from others: 0.82 without, 0.95 among straightliners. Subjective overuse: 0.83 without, 0.90 among straightliners. Social obligations: 0.85 without, 0.91 among straightliners. Source of concern: 0.64 without, 0.90 among straightliners. Source: Skogen and colleagues, Straightlining prevalence across domains of social media use, Scientific Reports, 2025, Table 2, from the LifeOnSoMe study of 3,285 adolescents aged 16 and over in Bergen, Norway, answering 64 items with five response options; straightliners were 5.4 percent of the sample and gave 16 or more identical answers in a row. A closing line says the weakest scale among attentive respondents looked strong among those who stopped reading, and notes that in this sample the overall conclusions barely moved, as the authors say. None of the numbers is a Cassi.ai measurement.

Identical answers are perfectly consistent, so straightlining pushes reliability up. A scale that looks stronger can be a scale fewer people read.

The format of the question matters as much as its position. The Survey Practice study by Richards and colleagues analyzed 14,613 web responses to the World Trade Center Health Registry survey in 2015. On screens narrower than 760 pixels, the five grids were shown in a stacked format, one row at a time. In all five grids, respondents were significantly more likely to straightline in the traditional grid than in the stacked one. Leaving items blank was also more likely in the traditional grid for four of the five. The authors could not assign devices at random, and say so, yet they still suggest a stacked variant for every screen size.

Why cleaning after fieldwork does not get the data back

UNIQUE INSIGHT Once the data is in, a straightliner leaves three options, and none of them recovers what was lost.

Keep the respondent, and the reliability of the scale and the correlations between scales move, as the LifeOnSoMe table shows. Drop the respondent, and the base shrinks in a way that is not random, because fatigue hits some profiles harder than others. Replace the respondent, and someone pays for extra days of fieldwork to buy completes that the questionnaire spent.

Dropping is also less clean than it looks. OpinionX separates four causes of straightlining: fraud, fatigue, items that do not apply to the respondent, and lists that never force a choice. Only the first is junk. An honest "never" down a list of services the respondent does not use is a straight line too, and a cleaning rule that deletes it removes a real answer.

So the decision belongs before fieldwork, with whoever signs the questionnaire. The researcher answers for the data quality the client receives. Today that researcher still pays for fatigue three times: in a weaker second half of the questionnaire, in hours of cleaning, and in extra fieldwork days to replace what was dropped.

Change the format and keep the results table

The fix that the evidence supports is to change how the items are presented while keeping what they measure. The respondent sees a different screen. The analyst receives the same variables, the same codes and the same tables.

The last row is honest about its evidence. A designed format keeps the measurement and changes the experience, and its effect on straightlining still has to be measured study by study.

1. List every grid in the questionnaire with its number of rows, its scale and its position.
2. Mark the grids that are validated scales or tracker items that must stay comparable with previous waves.
3. Move the grid the client cares about most earlier, because position alone changes the answers.
4. Present long grids one item per screen, with the same scale and the same labels.
5. Keep the variable names and answer codes identical, so the tabulation does not change.
6. Write the straightlining flag before fieldwork, as the longest run of identical answers and a threshold.
7. Read that flag in the soft launch data file, before the full sample is invited.
8. Report the share of flagged respondents in the methodology section of the report.

Change the screen, keep the table

Three schematic panels connected by arrows. Panel one, traditional grid: every item on one screen, drawn as six rows of five answer boxes with the same column marked in every row, labeled same column all the way down. Panel two, stacked items: one item per screen, drawn as a single item with its five answer options listed below, labeled each item gets its own answer. Panel three, results table: same variables and same codes, labeled tabulation does not change. Source line: Richards, Powell, Murphy, Nguyen and Yu, Gridlocked: the impact of adapting survey grids for smartphones, Survey Practice, 2016, with 14,613 web responses to the World Trade Center Health Registry survey in 2015; in all five grids, straightlining was significantly more likely in the traditional format than in the stacked one; devices were not randomly assigned. The panels are schematic and contain no data. A closing band says the respondent gets a different screen and the analyst gets the same tables as last wave.

Presentation changes what the respondent does. Variable names and codes decide whether the tables still line up with the previous wave.

Step five is the one that makes the change safe for a tracker. The respondent gets a new screen, and the tables still line up with last year.

Where Cassi.ai comes in

@Cassi.ai is a software engineering company specialized in the pains of market research, innovation and insights, founded by Cassiano Albuquerque after twenty years in market research.

Survey Studio is a Cassi.ai product that runs immersive research, giving every study its own format, such as gamified missions or a narrative in scenes, on one engine with a live fieldwork cockpit.

Survey Studio is our product for long online questionnaires. Each study gets a format designed for it from the brief and the study design. A brand diary can become missions with levels, a packaging test a store aisle with hotspots that open questions, a diagnosis a deck of cards or a board. Underneath every format runs one engine with the questionnaire's question types and its conditional logic, validated before fieldwork, and every format feeds the same results board. Each format is set up per project.

That is the last row of the table, built as software. The respondent moves through a task, and the analyst receives the variables the questionnaire defined. The logic check before fieldwork is the same discipline as testing the survey script before fieldwork, and the results board follows the principle in building the dashboard from the decision.

FAQ

What is survey fatigue?

The drop in attention and effort as a respondent moves through a long questionnaire. In a randomized web experiment, later questions got faster, shorter and more uniform answers.

What is straightlining in a survey?

Giving the same answer to every item in a series, or to an unusually long run of them. The LifeOnSoMe study flagged 16 or more identical answers in a row.

Should straightliners be removed from the data?

Not automatically. Some straight lines are honest answers to items that do not apply, as OpinionX points out. Flag them with a rule written before fieldwork and report the share.

Are grid questions bad for mobile respondents?

A Survey Practice study found more straightlining in traditional grids than in a stacked one-item-per-screen format, in all five grids tested.

Before the next long questionnaire

UNIQUE INSIGHT A respondent who straightlines a grid is still counted as a complete. The data looks whole, the scales look reliable, and the attention that made those answers worth collecting was spent on the screens before.

If your next study has a grid in its second half, count its rows and its position before it goes to programming. If you want to see what that block looks like in a format built for your study, send us the questionnaire your respondents never finish and we will design the format and show the board it feeds.

About the author

Cassiano Albuquerque is the founder of @Cassi.ai. He spent twenty years in market research operations before building software for research teams. To talk about questionnaire design for your studies, contact him on LinkedIn.

Published by Cassi.ai. Read the full article at https://www.cassiai.com/blog/survey-fatigue-grid-questions.