Atlas › Reading words and numbers

Case study 174 of 198

Does 'a few' grow with the crowd?

Does 'a few', 'several' or 'many' mean a bigger number to Jev when the thing counted is bigger (a stadium crowd vs a dinner party, grains of rice vs years)?

result15 questions

Asked how many "a few", "several" and "many" mean in five different settings, Jev scales only "many" to the thing being counted. "A few" is 3 whether it's people at a dinner party, a stadium crowd, emails, years or grains of rice; "several" stays at 3 or 4. "Many" runs from 6 to 7 (years) to 251 to 500 (people at a stadium).

dinneryearsemailsstadiumrice
Several34334
Many16-256-711-15251-500101-250
A few33333

How to read this: A grid: one row per word, one column per setting, each cell the number range Jev picked. A row that stays the same across the columns is a word that doesn't scale.

3 words x 5 settings.

In short

  • To Jev, "a few" means 3 whether it's dinner guests, a stadium crowd, emails, grains of rice or years.
  • Jev scales only "many", from 6 to 7 years up to 251 to 500 people at a stadium.
  • So Jev's "a few people at the stadium" is probably smaller than a person's would be.

What the data shows

reading words and numbers
Woody and Buzz Lightyear Everywhere Widescreen meme: "a few"; it's 3. it's always 3"a few"it's 3. it's always 3
How funny is this meme? Jev: 3/5, funny10%220%374%46%50%
  • "A few" is always 3: at a dinner, at a stadium, in an inbox, on the floor, in years.
  • "Several" barely moves: 3 at dinner, at the stadium and in the inbox; 4 for rice and years.
  • "Many" scales strongly: 6 to 7 years, 11 to 15 emails, 16 to 25 dinner guests, 101 to 250 grains of rice, 251 to 500 people at a stadium.

What it means, and what it doesn't

Jev treats "a few" and "several" as fixed small numbers, and "many" as relative to the situation. A person would likely scale all three to some degree, so Jev's "a few people at the stadium" is likely smaller than a person's. For "many", it reads the situation well.

Without human answers this shows Jev's pattern, not a gap with people, and 15 answers is a small grid. For the words without any setting, see "How many is 'a few'?".

Caveats

  • No human comparison. People weren't asked these exact questions, so there's no saying how much "a few" should grow at a stadium. It's a reasonable expectation that it grows; how much is open.
  • The project's settings and wording. The five sentences were written for this project. A stadium "when the gates opened" and grains of rice that "fell on the floor" are its choices; other wording might push the numbers.
  • Bins. Answers come in bins that widen as they go up (a single number up to 5, then 6 to 7, 8 to 10, and so on), so "a few" could shift a little within the "3" bin without showing.
  • Only three words. Three words in five settings is 15 answers. "Several" moving from 3 to 4 is a single step, which could easily have gone the other way.

Jev on this experiment

Would a person find it interesting to read?
Yes71%
Does it describe you?
Yes52%
Would you have predicted it?
No64%
How fair is the comparison?
The comparison is shaky
How much should a reader rely on it?
A little
Which caveat matters most?
No human comparison53%

Why ask this

"A few people came to my party" and "a few people were at the stadium" don't mean the same number. Readers scale vague amounts to what's being counted: many grains of rice is far more than many years. Whether a model does the same is a quick test of whether it reads the words or the situation they describe.

It matters when a model turns vague reports into numbers, in summaries, incident notes or estimates: "a few fans stayed behind" at a stadium and "a few guests stayed behind" at a dinner are probably not the same count, and a summary that turns both into 3 has quietly changed the report.

How this was done

The people and the data

This experiment has no human data; it compares Jev with itself across settings. The three words ("a few", "several", "many") and the 15 answer ranges come from the Reddit survey behind "How many is 'a few'?". The five settings were written for this project and chosen to differ in scale by orders of magnitude: people at a dinner party, people at a stadium when the gates opened, grains of rice that fell on the floor, emails in a day, and years ago.

What Jev was asked

Each word in each setting, answered with the same 15 ranges as the survey question:

Someone says "Several people came to my dinner party." How many people is that?

1 · 2 · 3 · 4 · 5 · 6 to 7 · 8 to 10 · 11 to 15 · 16 to 25 · 26 to 50 · 51 to 100 · 101 to 250 · 251 to 500 · 501 to 1,000 · More than 1,000

That's 15 questions, each with the ranges in three shuffled orders, averaged.

How it was measured

For each word and setting, the range holding the middle of Jev's answer. A word "scales" if that range moves with the size of the thing counted.

Where these questions live

15 questions across 1 topic of the map. Each opens on the map with every question in it.

Every question

All 15 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer
    Showing 0 of 15