The Empathy Spectrum
Opens the original study.
- Study
- Psychometric research, survey based
- Instrument
- Toronto Empathy Questionnaire, 16 items
- Context
- Human Factors in Interaction Design, DAU
- Method
- 14 statistical methods
- Led to
- MedAssist
01 Question
How empathetic are people, and does it vary with who they are?
The Toronto Empathy Questionnaire scores empathy from 0 to 64 across 16 statements, each rated from Never to Always. 277 people completed it.
- Responses277
- Statements16
- Top score64
02 Method
17 incomplete responses were removed, and 14 methods analysed the other 260.
The analysis moved from reliability and factor structure to item response theory, measurement invariance, a network of items, clustering and eight hypothesis tests.
- Removed17
- Analysed260
- Methods14
03 Findings
Empathy has two dimensions, and respondents fall into two profiles.
A single-factor model fits poorly. Caring items load on one factor and inhibition items on another. Latent profile analysis then separates the Resonators, who score higher on every item, from the Pragmatists.
- Resonators56%
- Pragmatists44%
- Factors2
03 Findings
Clustering reveals a small, low-reactive third group.
Clustering on a projection isolates 31 respondents, about 12%, with a mean score near 32. The number of profiles depends on the method: latent profile analysis finds two and clustering finds three.
- Respondents31
- Of the sample12%
- Mean score32
03 Findings
Gender relates to empathy score. Age does not.
Women scored about four points higher (r = 0.287). Age showed no relationship, partly because most respondents were 20 to 25. Twelve of the 16 items also work differently for men and women, so a single score masks real differences.
- Gender and scorer = 0.29
- Items differ by gender12 of 16
- Of the sample was male57%
Respondents
277 people
Each dot is one respondent. Dots show group sizes, not individual answers.
04 Directions
- The Detachedabout 32
- The Pragmatist37.0
- Sample mean43.5
- Published normabout 47
- The Resonator48.6
- 56% of the sampleThe ResonatorScores high on every item. Responds to emotional narrative, community and visible impact.
- 44% of the sampleThe PragmatistPrefers clear hierarchy and rational framing. Emotional content works when the user opts in.
- About 12%The DetachedWants utility first: factual calls to action with minimal emotional friction.
- 01Empathy-adaptive layeringTone, information density and interaction patterns adapt to the user's profile.
- 02Design for the core drivers firstFeatures that touch the most central items spread through the whole empathy network.
- 03Gender-neutral defaults, adjustable toneBecause items work differently by gender, the defaults stay neutral and the tone stays adjustable.
- 04No guilt for the low-reactiveUtility-first design removes emotional pressure for the Detached.
05 Limits
- The sample is mostly university students aged 20 to 25, and 57% are male.
- The scale shows measurement problems: a one-factor model fits poorly and many items differ by gender.
- The profiles depend on the method. Latent profile analysis gives two and clustering on a projection gives three.
The research site
Get in Touch