THIS
OR THAT

Two options are easy because the cost of deciding grows with the logarithm of how many options there are, not with the options themselves. W. E. Hick fitted that curve to his own data in 1952, and on it a ten-way choice does not cost five times what a two-way choice costs, only about twice.

The less obvious part is that a forced choice is not just faster. It is also harder to fudge, which is why measurement has been built out of two-item pairs for a century, and why the pairs that work are engineered rather than picked.

Key Takeaways

• Hick found choice reaction time fits the function log of n plus one, where n is the number of alternatives. On his own fitted curve two alternatives take about a quarter of a second and eight take about half.
• Louis Leon Thurstone built a measurement method on pairs in 1927, and his rule was that "no equal judgments are allowed". No ties, no both, no neither.
• The method turns a heap of binary answers into a scale, but it costs half of n times n minus one comparisons. Five items is ten pairs; twenty items is 190.
• Forced-choice personality tests inflate by about 0.06 standard deviations under pressure, but the same meta-analysis found the advantage over ordinary rating scales is trait-by-trait rather than general.
• Matching the two options for social desirability is what does the work. Unbalanced pairs inflated by 0.25 standard deviations; balanced ones by 0.03.

Why is a forced choice between two things so easy to answer?

Because the cost of deciding follows a logarithm, and logarithms flatten fast.

W. E. Hick published the result in the Quarterly Journal of Experimental Psychology in 1952, working at the Medical Research Council's Applied Psychology Research Unit in Cambridge. His abstract describes two experiments, the first "a conventional choice-reaction experiment, with various numbers of alternatives up to ten, and with a negligible proportion of errors". The first of them, he writes, "was carried out mainly to confirm the fitness of the function" that gives choice time as a constant times the logarithm of n plus one. It did.

The curve he fitted to his own readings was 0.518 times the base-ten logarithm of n plus one, in seconds, where n is the number of alternatives. He fitted 0.626 times the same function to Merkel's much older 1885 data. Put two alternatives through Hick's curve and you get about a quarter of a second. Put eight through it and you get about half. Four times the options, roughly twice the time.

That is the shape worth carrying around. Adding options is expensive when there are few and cheap when there are many, so the jump from one option to two costs you far more than the jump from eight to nine. A pair is about as close to free as a real decision gets, and it is why the reaction time of a two-alternative key task is measured in a couple of hundred milliseconds. Hick's own n of one, a single stimulus with a single response, is what the literature calls simple reaction time, and his curve puts it lower again.

Hick's own headline was different and stranger: the rate at which people take in information is "on the average, constant with respect to time, within the duration of one perceptual-motor act, and has a value of the order of five 'bits' per second". The log curve falls out of that.

Robert Proctor and Darryl Schneider, reviewing the work since Hick in the same journal in 2018, call what Hick established "one of the few widely acknowledged laws in psychology" and note it "is regarded as a fundamental law of interface design for human-computer interaction". They also list where the simple form needs qualifying: "effects of stimulus-response compatibility, practice, very large set sizes and sequential dependencies". A strong regularity, then, not a constant of nature.

How does forced choice turn opinions into measurements?

By throwing away the option of not choosing, and then counting.

In 1927 Louis Leon Thurstone published a paper in Psychological Review setting out what he called the law of comparative judgment. Its object, he wrote, was "to describe a new psycho-physical law which may be called the law of comparative judgment and to show some of its special applications in the measurement of psychological values".

The method underneath it is the method of paired comparison, and the setup is exactly a this or that round. You are shown two things at a time and asked which has more of some attribute. Thurstone was deliberately loose about what the attribute could be: "it may be gray values, or weights, or excellence, or any other quantitative or qualitative attribute about which we can think 'more' or 'less' for each specimen."

The constraint is the interesting bit. His law applies, he says, to an observer who "compares a series of stimuli by the method of paired comparison when no 'equal' judgments are allowed". No ties. Not "they are about the same", not "both", not "it depends on the day". You pick one, every time, and the proportion of times each item beats each other item becomes the raw material for a scale.

Thurstone was honest about what that scale is. "The psychological scale is at best an artificial construct," he wrote. "If it has any physical reality we certainly have not the remotest idea what it may be like." What the method gives you is not a reading off some hidden dial. It is a consistent ordering, built entirely out of answers to a question nobody could have answered directly.

He saw the wider use immediately. The same paper says the law "has been applied in the measurement of such psychological values as a series of opinions on disputed public issues", adding that "the latter application of the law will be illustrated in a forthcoming study". Pairs of options, in other words, put to people who cannot otherwise tell you how strongly they feel.

Does a forced choice get more honest answers?

Sometimes, and the caveats are more useful than the headline.

Ask someone to rate a statement about themselves from one to five and they can inflate every answer. Ask them to pick which of two statements is more like them and the inflation has nowhere to go. That is the theory, and it has been tested at scale in the least romantic corner of psychology, which is job selection testing.

Mengyang Cao and Fritz Drasgow published a meta-analysis in the Journal of Applied Psychology in 2019 comparing forced-choice personality scores between low-stakes settings and high-stakes ones, meaning tests people had a real reason to game. They define the effect as the standardised difference in scores between the two situations, and report that "the overall score inflation effect size for FC personality measures is 0.06", which is six hundredths of a standard deviation and close to nothing.

Then the complication, which their abstract soft-pedals and their results section does not. Set against published figures for ordinary single-statement scales, the forced-choice advantage is trait-by-trait rather than general. In between-subjects studies they found forced choice significantly lower on emotional stability, 0.37 against 0.64, on openness, 0.11 against 0.65, and on agreeableness, 0.00 against 0.48. In within-subjects studies the gap on emotional stability was wider still, 0.20 against 0.93, and on openness 0.25 against 0.76.

But forced choice came out higher on extraversion, 0.31 against 0.16 between subjects, and higher on agreeableness within subjects, 0.75 against 0.47. Their own verdict on that hypothesis is "mixed support". Note also that most of these trait figures sit above the pooled 0.06. The pooled figure is a weighted average over every trait and study in the analysis, so it is not the same quantity as any single trait-by-design comparison, and it should not be read as a per-trait floor.

What does hold up cleanly is the condition. Forced-choice scales, they report, "were consistently found to be more faking-resistant when constructed with statements balanced in social desirability", and their moderator table puts numbers on it: pairs with no desirability balance inflated by 0.25 standard deviations, pairs with it by 0.03.

So the format does not buy honesty. Matching the two options for social desirability buys honesty, and the format is what lets you do the matching.

Property	Two matched statements	One statement, rated one to five
What the respondent can do	rank two things they cannot both claim	move every answer in the flattering direction
Measured inflation under pressure	0.06 overall; 0.00 to 0.75 across the traits quoted above	0.16 to 0.93 across the same traits
What it depends on	the two options costing the same socially	nothing the writer controls
Where the format loses	higher on extraversion, 0.31 against 0.16	lower on extraversion in the same comparison
What it costs to build	a matched pair for every item	one sentence per item

What makes a good this or that pair?

Four rules, and every one of them comes out of the measurement literature rather than out of taste.

1. Match the two options for social desirability. This is the single finding that separates a pair that measures something from a pair that does not: 0.25 against 0.03 in Cao and Drasgow's own table. If one side is the answer that makes you look better, everyone takes it and you have learned nothing.
2. Count your pairs before you write them. Thurstone's method needs half of n times n minus one comparisons for n items. "For a scale of five specimens there will be ten observation equations," he writes. Ten items is 45 pairs, twenty is 190, and that growth is why paired comparison is used on short lists and almost never on long ones.
3. Forbid the tie, out loud. Thurstone's law is derived for the case where no equal judgments are allowed, and the practical reason follows from the same logic: once "both" is on the table it costs nothing to pick, and an answer that costs nothing produces nothing to count.
4. Do not assume every pair is equally easy. Thurstone's simplest formulation, his Case V, assumes every item has the same discriminal dispersion, the same spread of momentary impressions, and comes with his own warning that the "assumption should not be made without experimental test". That is a narrower assumption than it sounds, but the practical lesson is broader: a set where three pairs are agonising and twelve are obvious is not a scale, it is three questions and twelve pieces of filler.

None of which makes the pair a neutral instrument. Somebody chose the two things, and that choice fixes the entire question. The reason a good this or that round feels like a fair fight is that somebody did the boring work of making it one.

For a forced choice that costs the chooser something, Would You Rather runs the same mechanic with the stakes turned up, and the collection of quizzes and games has the rest of the formats.

Frequently asked questions

What is a forced-choice question?

A question that offers a fixed set of options, usually two, and does not allow a neutral, tied or blank answer. It is standard in psychological measurement, where the more familiar rating scale lets a respondent sit in the middle of every item. The forced version removes that hiding place, which is the point of it: an answer that costs nothing to give tells you very little.

Why do personality tests use pairs instead of rating scales?

To make the test harder to game when the stakes are high. A 2019 meta-analysis in the Journal of Applied Psychology put overall score inflation for forced-choice personality measures at 0.06 standard deviations between low-stakes and high-stakes conditions. The same paper found the advantage over single-statement scales is uneven across traits rather than universal, and that it depends heavily on whether the two statements were matched for social desirability. This is evidence about personality testing in personnel selection, not about surveys in general.

Does having fewer options really make people decide faster?

Yes, and the relationship is logarithmic rather than proportional. W. E. Hick fitted choice reaction time to the function log of n plus one in 1952, where n is the number of alternatives, and a 2018 review in the same journal calls it one of psychology's few widely acknowledged laws. Practically, that means the first option you add is expensive and the tenth is cheap, so two is near the floor. The reviewers note the simple form needs qualifying for practice, very large sets and other factors.

What did Thurstone add to paired comparison?

The machinery for turning wins into distances. Comparing things two at a time long predates him, and he presented his law as continuous with earlier psychophysics rather than as a break from it: it "is implied in Weber's law and in Fechner's law", he wrote, and he described it as "basic for all experimental work" on both. What 1927 supplied was a way to get from the proportion of times each item beats each other item to a scale with spacing on it, rather than just an order.

Why does a this or that round die when one answer is obvious?

Because an obvious answer carries no information, and the measurement literature says the same thing with numbers attached. Pairs whose two sides are not matched for desirability inflated eight times as much as matched ones in the 2019 meta-analysis, precisely because everybody drifted to the same side. A pair that everyone answers identically has measured the pair, not the people. That is also why a game like Higher or Lower, where one side is simply larger, is a quiz rather than a preference test.

Sources

• W. E. Hick, On the Rate of Gain of Information, Quarterly Journal of Experimental Psychology vol. 4 no. 1 (1952), pages 11 to 26. The article of record is in the Quarterly Journal of Experimental Psychology; the copy read here was a full scan of a university course PDF reached through an Internet Archive capture, used because the publisher's page is not reachable from this machine. The abstract on page 11 supplies the experiment description and the five-bits-per-second finding; the fitted functions and the Merkel comparison are from pages 14 and 15. Retrieved 10 August 2026, scan of Hick 1952
• Robert W. Proctor and Darryl W. Schneider, Hick's Law for Choice Reaction Time: A Review, Quarterly Journal of Experimental Psychology vol. 71 no. 6 (2018), pages 1281 to 1299. Abstract read in the PubMed record; it is the source for the "widely acknowledged laws" phrasing, the interface-design remark and the list of qualifications. Retrieved 10 August 2026, PubMed record 28434379
• Louis Leon Thurstone, A Law of Comparative Judgment, Psychological Review vol. 34 (1927), pages 273 to 286. Full text read in the Mead Project transcription hosted by Brock University, which reproduces it with the American Psychological Association's permission. Brock's live page is behind a bot wall from this machine, so it was read through an Internet Archive capture. Retrieved 10 August 2026, Mead Project transcription of Thurstone
• Mengyang Cao and Fritz Drasgow, Does Forcing Reduce Faking? A Meta-Analytic Review of Forced-Choice Personality Measures in High-Stakes Situations, Journal of Applied Psychology vol. 104 no. 11 (2019), pages 1347 to 1368. Full PDF read. The overall figure and the desirability-balance condition are from the abstract on page 1347; the trait-by-trait comparison against single-statement measures is from page 1356; the desirability-balance moderator values are from Table 3 on page 1355. The article of record is in the Journal of Applied Psychology; the copy read here was a full PDF rehosted on a personal site, used because the publisher's own page is not reachable from this machine. Retrieved 10 August 2026, full PDF of Cao and Drasgow