Atlas › Memory and maps

Case study 22 of 198

Jev's mental map puts Europe too far south

Asked which of two cities on different continents is farther north, does Jev share people's classic error of placing Europe too far south of North America?

result362 questions

Asked which of a European and a North American city is farther north, Jev picks the North American one 89% of the time, though the European city is the northern one in 66% of the pairs. It's right 16% of the time when Europe is north and 100% when it's south. Like people's, its mental map puts Europe too far south: it's 97% sure Ottawa is farther north than Chisinau, Moldova, though Ottawa lies 1.6 degrees farther south. For two US cities it's right 89% of the time.

Europe vs North America44%
other regions59%
US vs US89%
Europe-NA, Europe north16%
Europe-NA, Europe south100%

How to read this: Bars show the share Jev gets right in each set. The key pair is the last two: Europe-North America pairs split by which city is really farther north. A fair judge would do about as well on both.

122 Europe-North America pairs, 362 pairs in all; 90% interval on the Europe-North America share right [0.377, 0.516]. Not an order effect here: Jev picks the city named first in 51% of the Europe-North America pairs. It is elsewhere: on pairs from other regions it is right 80% when the northern city is named first and 35% when it is named second (recall_mental_map_west). The human pattern is from Friedman & Brown 2000; there are no human answers to these pairs. Surest wrong: Ottawa, Canada vs Chisinau, Moldova (97% on the wrong city); Montreal, Canada vs Chisinau, Moldova (96% on the wrong city).

In short

  • Jev shares a classic human distortion: it imagines Europe too far south, naming the North American city as farther north in 89% of mixed pairs.
  • The error is one-sided: Jev is right on every pair where the North American city really is north, and on 16% where the European one is.
  • No people answered these exact pairs; the human side is a published pattern. Within the US, Jev is right 89% of the time.

What the data shows

memory and maps
Simba Shadowy Place meme: Jev: North America is farther north; what about Paris?Jev: North America is farther northwhat about Paris?
How funny is this meme? Jev: 2/5, slightly funny12%251%346%41%50%
  • It almost always picks North America. In Europe-North America pairs, Jev names the North American city as farther north 89% of the time, though the European city is actually farther north in 66% of these pairs.
  • So it's wrong exactly where the published human error predicts. When Europe is north, Jev is right 16% of the time; when Europe is south, 100%.
  • And sure of it. It's 97% sure Ottawa is north of Chisinau, Moldova, and 96% sure Montreal is. Ottawa lies 1.6 degrees south of Chisinau, Montreal about 1.5.
  • It isn't about word order here: Jev picks the city named first in 51% of these pairs.
  • The US control is fine: two US cities, right 89% of the time. Pairs from other regions are harder: 59%.

What it means, and what it doesn't

Jev's mental map has the same tilt as people's: Europe slides south. One plausible reason is that text about cities rarely states their latitude, while everyday descriptions ("mild Mediterranean winters", "freezing Canadian winters") point the wrong way. A model learning geography from words would pick up the impression, not the coordinates.

It doesn't mean Jev gets geography wrong in general: within the US it's right 89% of the time. The error is specific to comparing continents, where impressions stand in for numbers.

Caveats

  • No human answers to these pairs. The comparison with people is with a published pattern (people imagine Europe well south of where it is, Friedman and Brown, 2000), not with people answering these exact questions. No item-level human data was found.
  • The pairs were chosen to test the error. The pairs are European and North American cities 0.5 to 6 degrees of latitude apart, where the error bites. Two thirds of those pairs have Europe north. Across every possible pair the error would matter less.
  • "Well-known" by population. Cities were chosen by size (over 1.5 million people, or national capitals over 300,000) from the GeoNames database, so some are less familiar to English speakers than others.

Jev on this experiment

Would a person find it interesting to read?
Yes75%
Does it describe you?
No57%
Would you have predicted it?
No64%
How fair is the comparison?
The comparison is shaky
How much should a reader rely on it?
Moderately
Which caveat matters most?
No human answers to these pairs55%

Why ask this

Is Rome north or south of New York? People tend to guess south; it's actually slightly north. People's mental maps line Europe up with the US, when Europe sits well to the north; the psychologists Friedman and Brown documented the pattern in 2000.

A model knows maps only through text. Whether it inherits the human distortion or the coordinates is a small window on how it stores geography, and it matters for anyone asking it about climate, daylight or travel between the two continents, where "how far north" is the whole question.

How this was done

The people and the data

The truth comes from coordinates in GeoNames, an open geographic database. The study built three sets of pairs of well-known cities (over 1.5 million people, or national capitals over 300,000): 122 pairs of one European and one North American city between 0.5 and 6 degrees of latitude apart; 150 pairs from other regions of the world; and 90 pairs of US cities as a control. There are no human answers to these pairs; the human side is the published pattern.

What Jev was asked

Which city is farther north: Ottawa, Canada or Munich, Germany?

Answers: Ottawa, Canada · Munich, Germany

Each question was also asked with the two answers in the other order, and the answers averaged.

How it was measured

The share of pairs Jev gets right in each set. For the Europe-North America pairs, how often it picks the North American city, and its accuracy split by which city is really farther north: the human error predicts misses exactly when Europe is north. A check that the lean isn't just a preference for the city named first.

Where these questions live

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

Every question

All 362 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 362