Table 1 Gaze metrics indexing processes that may support improvements in reasoning

From: Eye gaze patterns reveal how reasoning skills improve with experience

Cognitive process

Gaze metric

Evidence supporting H1 BF10 ≈ P(H1 | data)/P(H0 | data)

H1 = POST < PRE

H0 = Group + Time

Reasoning

Comprehension

H1 = Group × Time BF10

  

PRE Mdn [95% CI]

POST Mdn [95% CI]

BF10 (% error)

PRE Mdn [95% CI]

POST Mdn [95% CI]

BF10 (% error)

BF10 (% error)

Visual search

Decrease in the number of fixations on any scale before homing in on the relevant scales

5.75 [4, 7]

5 [4, 5]

6.51* (±<0.00)

5 [4, 5.75]

4 [3, 5]

0.67 (±<0.00)

0.60 (±2.79)

Relational thinking

Decrease in the total duration of fixations on relevant relations after homing in on the relevant scales

2526.5 [1010.50, 3259.00]

1122.5 [555.99, 1854.50]

240.03*** (±<0.00)

1635.75 [1069, 2256.25]

1725 [1387.5, 1990.18]

0.30˙ (±0.02)

3.66* (±1.98)

Relational integration

Fewer saccades between the two relevant scales after homing in on the relevant scales

3 [1.75, 3.75]

2 [1, 3]

3.15* (±<0.00)

2 [2, 3]

2.5 [2, 3]

0.14˙ (±<0.00)

1.58 (±2.62)

  1. Mdn [95% CI]: Median with 95% confidence intervals, calculated with 1000 bootstrap iterations. H1 = POST < PRE assessed with Bayesian paired single-sided t test. Interaction models tested with Bayesian mixed regressions. Estimations made using BayesFactor’s53 default Cauchy prior scale \(r = \frac{1}{2}\sqrt 2\) and prior uniform probability to the models. Refer to Table S1 for specification of the models and posterior odd estimates. Approximate classification scheme for the interpretation of Bayes factors from ref. 40: ***Extreme evidence H1, *Moderate evidence H1 and ˙Moderate evidence for H0