Empathising–Systemising Theory

A contested psychological theory that profiles separately measured empathising and systemising dimensions, classifies their relative balance, and tests group-average hypotheses about sex-coded samples and autism.

Version

v2 · 2026-09-06 · History

Domain-specific #

1762

Origin domain

psychology

Subdomain

individual differences and autism research

Aliases

Empathizing–Systemizing theory, Empathising-Systemising theory, E–S theory (psychology)

Core Idea

Empathising–Systemising Theory (E–S theory) is a named and contested psychological theory of individual differences. It proposes two separately measured dimensions. Empathising is defined within the theory as both identifying another person's mental state and having an affective orientation toward responding to it. Systemising is defined as the drive to analyze or construct rule-governed systems by detecting regular relations—often represented as input–operation–output or “if p, then q” rules.^[1][2] The theory then compares a person's standardized empathising score E with a standardized systemising score S, commonly through a difference profile D = S - E, and assigns sample-relative profile labels such as Type E, Type S, Type B (balanced), Extreme E, or Extreme S.^[3]

That score-and-classify operation is the abstraction's stable identity. It allows researchers to state testable hypotheses about distributions of profiles across samples. Foundational formulations proposed average differences between male- and female-coded groups and extended the model to autism through the extreme male brain (EMB) hypothesis.^[1] These are empirical claims made by the theory, not facts built into this encyclopedia entry. Group means do not classify every member of a group, profile labels are not anatomical brain types, and neither an E–S score nor an “Extreme S” classification diagnoses autism.

The node therefore preserves a theory as an object of inquiry: its constructs, measurements, transformations, predictions, and failure conditions. Acceptance into the encyclopedia recognizes a durable reasoning structure in psychology. It does not endorse binary sex essentialism, a universal empathy deficit in autism, a prenatal-hormone explanation, or any stronger causal interpretation.

Structural Signature

The recurring structure is:

defined empathising construct + defined systemising construct → separate operational measures → standardized within-person comparison → sample-relative profile classification → group-distribution hypotheses → empirical testing and revision.

Seven roles are load-bearing:

1. Two construct definitions. Empathising combines mental-state attribution and responsive affect in the theory; systemising concerns rule extraction or construction. Ordinary sympathy and ordinary interest in systems are not automatically the constructs.
2. Separate observations. The Empathy Quotient (EQ) and Systemizing Quotient or revised Systemizing Quotient (SQ/SQ-R) are common self-report operationalizations, not the constructs themselves.^[4][2]
3. A common comparison scale. Raw scores require a declared reference sample, form, normalization, and scoring convention before subtraction is meaningful.
4. A relative profile. In a prominent large-sample implementation, standardized S and E scores produce D = S - E. The sign and magnitude express relative balance, not absolute competence.^[3]
5. Classification thresholds. Cut points partition the reference distribution into profile categories. Their meaning depends on the norming procedure; they are not discovered anatomical kinds.
6. Population-level predictions. The theory predicts differences in profile distributions among specified groups. Valid tests require comparable measurement and honest attention to overlapping distributions.
7. A contested explanatory extension. EMB relates autism to a population shift toward Type S or Extreme S. That extension must be distinguished from the core two-dimension profile and tested rather than assumed.

The invariant is: separately operationalized E and S are placed on a declared common scale, their relative balance is classified, and the resulting profile distribution is used to formulate falsifiable individual-difference or group-average claims. Removing either dimension yields a one-trait account. Removing standardization makes the difference score uninterpretable. Removing the comparison and classification leaves two questionnaires, not E–S theory.

What It Is Not

E–S theory is not empathy generally. Empathy is a family of cognitive, affective, interpersonal, and situational phenomena. Meta-analytic estimates of autistic–nonautistic differences change substantially with whether a measure targets cognitive, affective, or undifferentiated empathy; recent evidence does not justify one global deficit claim.^[5]

It is not Theory of Mind. Mental-state inference is part of cognitive empathising in many formulations, but E–S adds affective response, systemising, relative-score profiles, and group hypotheses.

It is not Systems Thinking. Systems Thinking examines wholes, feedback, interdependence, and system behavior as a reasoning practice. Systemising in E–S is a proposed psychological disposition to seek or construct rules. Someone may practice Systems Thinking without receiving a high SQ score, and a high SQ score does not establish expertise in complex-system analysis.

It is not the dual-process contrast between System 1 and System 2. Those labels concern fast/automatic and slow/deliberative processing; the word system creates only a lexical resemblance.

It is not the EQ, SQ, SQ-R, or Autism-Spectrum Quotient (AQ). Those are instruments. Instrument reliability, factor structure, measurement invariance, and criterion validity must be evaluated separately. The AQ is also not a clinical diagnosis.

It is not a biological scan. “Brain type” in this literature is a psychometric profile label. No neuroanatomical type follows from a difference score. Nor is a male-coded group mean evidence that all men systemise more than all women, or that traits are caused by sex.

Finally, the core E–S profile is not exactly synonymous with EMB. EMB is a stronger autism-related extension. A researcher can use E, S, and D profiles while rejecting EMB's causal, developmental, or sex-linked interpretation.

Scope of Application

The theory belongs to individual-differences psychology, psychometrics, autism research, and research on sex- or gender-coded group comparisons. It is used to construct questionnaires, compare profile distributions, test correlations with other traits, and generate hypotheses about occupational interests, social cognition, rule learning, and autistic characteristics. Its recurrence includes the original SQ and EQ studies, replications and adaptations in other languages, studies that compare E–S profiles with broader personality factors, and large public-sample tests of enumerated predictions.^{[2][4][6][7][3]}

The scope is narrower than “explaining human cognition.” E–S constructs do not exhaust reasoning, personality, empathy, social interaction, or technical ability. Nettle found that EQ was closely related to Big Five agreeableness, while systemising had its own personality correlations; those relations challenge claims of complete novelty even if they do not erase the two-score framework.^[6]

Applications require exact population labels. A dataset that records binary sex cannot establish a claim about gender identity without additional measurement. A convenience web sample cannot be treated as representative of a nation. Greenberg and colleagues analyzed 671,606 participants recruited through a Channel 4 website, used short self-report forms, relied on self-report of autism diagnosis, and restricted the reported binary comparison after excluding other or prefer-not-to-say responses.^[3] Its scale supports precise estimates within that sample; it does not remove selection, reporting, cultural, or measurement limitations.

Autism applications require still more restraint. Autism is heterogeneous across language, cognition, sensory experience, social motivation, alexithymia, camouflaging, age, support needs, and co-occurring conditions. E–S profiles may be studied among autistic people, but they neither define autism nor decide whether one person is autistic. Reciprocal accounts such as the double-empathy problem also locate some social difficulty in mismatched perspectives between autistic and nonautistic people rather than a one-sided lack within the autistic person.^[8]

Clarity

A case is genuinely an E–S-theory use when five questions can be answered:

1. What definitions of empathising and systemising are being tested?
2. Which validated instrument versions produce the two scores?
3. Against which reference sample and transformations are those scores standardized?
4. How is relative balance computed and how are any profile thresholds set?
5. Which population-level prediction is tested, with what evidence capable of disconfirming it?

For example, Greenberg and colleagues used short-form measures and standardized them against a typical sample. In their implementation,

\[ S=\frac{SQ\text{-}R\text{-}10-\mu_S}{20},\qquad E=\frac{EQ\text{-}10-\mu_E}{20},\qquad D=S-E. \]

They classified the bottom 2.5% of the reference D distribution as Extreme E, the next 32.5% as Type E, the middle 30% as Type B, the next 32.5% as Type S, and the top 2.5% as Extreme S.^[3] Those percentages are set by cut points in the norming distribution. They do not show that nature contains five discrete clusters.

A profile is relative. A person can score high on both dimensions yet have D approximately zero, and another can score low on both yet also have D approximately zero. Both may be Type B by relative balance while differing greatly in absolute scores. Reporting only the type loses that information. Likewise, a group shift in mean D can coexist with extensive individual overlap, so group membership is a poor substitute for individual measurement.

Manages Complexity

The theory compresses a large and heterogeneous set of observations into a two-coordinate profile and a small taxonomy. That compression makes claims testable: instead of the vague statement that one group “thinks differently,” a study can declare instrument versions, estimate distributions of E, S, and D, compare effect sizes, and test whether profile frequencies shift as predicted.

The compression also creates risks. Two numbers cannot preserve every component of empathy or every form of rule learning. A difference score can hide high–high and low–low profiles. Thresholds can turn continuous variation into apparently categorical “types.” Self-report can reflect self-concept, social norms, item familiarity, and willingness to endorse traits as well as underlying ability. A model manages complexity responsibly only when it retains the underlying scores, uncertainty, sample definition, and measurement limitations alongside the profile label.

The measurement literature supplies a crucial guardrail. A COSMIN-based review found substantial evidence gaps across empathy measures and emphasized that absent measurement-invariance or differential-item-functioning evidence, observed group differences cannot automatically be read as equivalent trait differences.^[9] Reliability is necessary but not sufficient: a scale can consistently reproduce a score affected by item wording or group-specific interpretation.

Abstract Reasoning

E–S theory licenses conditional, not categorical, deductions. If two validated measures are comparable across groups, the same transformation is used, and a group distribution shifts on D, then the study supports a difference in measured relative profile for that sample. It does not by itself establish a causal pathway, biological essence, occupational aptitude, or individual prediction.

The structure separates four inferential levels:

1. Construct level: whether empathising and systemising are coherent and distinguishable psychological constructs.
2. Measurement level: whether EQ and SQ-family scores validly and comparably operationalize those constructs.
3. Classification level: whether the chosen normalization and thresholds produce a useful relative profile.
4. Explanatory level: whether differences by sex-coded group or autism status follow the theory's proposed explanation rather than sampling, culture, personality overlap, reporting, or another mechanism.

Evidence at one level does not automatically settle the next. A reproducible mean difference does not validate the profile categories; a reliable profile does not establish EMB; a correlation with autistic traits does not diagnose autism. Conversely, evidence against a strong sex-linked or causal claim need not erase the descriptive operation of measuring two dimensions and comparing their balance. This layered logic is why the theory remains an autonomous abstraction even while its strongest interpretations are contested.

Knowledge Transfer

The E–S framework transfers literally across studies only when the same roles survive: two defined constructs, separate comparable measures, a declared relative-score transformation, and empirical profile hypotheses. It has transferred from original UK adult samples into clinical and nonclinical studies, alternative questionnaire versions, translations, personality comparisons, and large online samples.^[7][3]

What transfers less securely are thresholds and interpretations. Norms derived from one instrument, language, age band, or recruitment channel should not be copied into another without validation. A translated EQ may preserve item meaning unevenly; an occupational sample may have restricted ranges; a short form may not reproduce the factors of the long form. The label “Type S” is therefore portable only with its scoring provenance.

Outside psychology, a two-axis relative classification may resemble E–S, but that skeleton belongs to more general abstractions such as Measurement and Classification. Calling an organization “empathising” and an engineering process “systemising” without validated psychological constructs is metaphor, not an instance of E–S theory. The domain-specific residue is the named construct pair, its instrument history, profile taxonomy, and disputed sex/autism hypotheses.

Examples

Reference-profile classification. Suppose a study administers specified EQ and SQ-R forms and converts each raw score using a validated reference sample. Participant A has S = 0.8 and E = 0.1, giving D = 0.7; Participant B has S = 1.2 and E = 1.1, giving D = 0.1. A may fall in a Type S band while B falls in Type B even though B's absolute systemising score is higher. This demonstrates that the taxonomy concerns relative balance, not a ranking of total intelligence or technical ability.

Large-sample prediction test. Greenberg and colleagues tested ten predictions in 671,606 records, including sex-coded average differences, profile distributions, and shifts associated with self-reported autism diagnosis.^[3] The dataset illustrates the complete theory operation: short-form measures, standardization, D, percentile-defined types, and group comparisons. It also illustrates why scale is not the same as generalizability: recruitment and self-report boundaries remain.

Measurement challenge. A researcher finds a lower EQ mean in one group. Before interpreting it as lower empathising, the researcher tests factor structure, invariance, and differential item functioning. If items function differently, the apparent group difference may partly reflect measurement rather than the intended construct. This case uses the theory critically rather than treating questionnaire output as transparent.

Autism boundary case. An autistic participant may be high in systemising and high in affective empathy while finding some mental-state inference tasks difficult. The person's scores cannot be collapsed into “no empathy,” and the profile does not supply a diagnosis or causal explanation. Separating cognitive from affective empathy is empirically important: a 2025 meta-analysis reported much larger autistic–nonautistic effects for cognitive or unidimensional measures than for affective empathy, with high-quality affective-empathy studies showing no significant difference.^[5]

Structural Tensions

Continuous dimensions versus discrete types. E, S, and D are continuous, but communication often uses five categories. Categories simplify comparisons; they also imply boundaries sharper than the data. A responsible use reports score distributions and threshold provenance.

Relative balance versus absolute level. D = S - E makes within-person asymmetry visible, yet erases whether both dimensions are high or low. Preserve E, S, and D together whenever absolute level matters.

Self-report versus performance. EQ and SQ-family instruments efficiently measure endorsed tendencies, but self-description is not identical to observed ability or behavior. Convergent evidence from tasks, informants, or real-world outcomes is needed for stronger claims.

Replicable group mean versus individual stereotyping. A statistically stable group difference may be scientifically informative while remaining useless or harmful as a deterministic rule about one person. Distributional overlap and uncertainty must accompany group summaries.

Descriptive profile versus causal explanation. Score patterns can recur even when prenatal hormones, socialization, educational exposure, personality, item interpretation, or selection provide competing explanations. The profile and its proposed cause are separable claims.

One-sided deficit versus reciprocal mismatch. E–S and EMB accounts can frame autistic social differences as an individual imbalance. Double-empathy accounts emphasize mutual misunderstanding across differently situated people.^[8] Study design must distinguish these levels rather than using one to dismiss the other by definition.

Construct breadth versus measurement specificity. “Empathy” is broader than any questionnaire. Recent measure-sensitive meta-analysis shows that conclusions depend strongly on whether studies isolate cognitive and affective components.^[5] Broad rhetoric cannot outrun the measured subconstruct.

Structural–Framed Character

The abstraction is mixed-framed, with a strong framed component (provisional framedness score: 0.60). Its formal spine—two scores, common-scale transformation, difference profile, thresholds, and distributional predictions—is structural and reproducible. The skeleton supports explicit calculations and falsification.

Its content, however, depends on humanly framed constructs and institutions: questionnaire language, reference populations, sex/gender coding, diagnostic practices, cultural expectations, and contested interpretations of social behavior. The labels “empathising,” “systemising,” “male brain,” and “female brain” carry evaluative and historical accents that the mathematics cannot neutralize. Changing a norming sample can change a type assignment without changing the person. Changing an item interpretation can change measured group differences without changing the intended trait.

The score reflects substantial human-practice dependence, moderate vocabulary dependence, and moderate institutional and measurement dependence. The framework is not merely rhetorical because the operations are explicit; it is not purely structural because construct validity and social interpretation are constitutive.

Structural Core vs. Domain Accent

The liftable structural core is measure two dimensions, place them on a common scale, compare their relative balance, partition a reference distribution, and test group-level predictions. That operation instantiates Classification and depends on Measurement. It can be represented without psychological vocabulary.

The domain accent is indispensable to E–S theory itself: the particular definitions of empathising and systemising; EQ/SQ instrument families; sample-normed “brain type” labels; the history of proposed sex-linked distributions; and the EMB extension to autism. Removing those elements yields a generic two-axis classifier already covered by broader catalog abstractions. Transferring the labels to organizations, machines, or cultures without psychological operationalization is analogy.

The domain-specific classification is therefore stable. The abstraction recurs across multiple psychological studies and instrument settings, but its identity does not travel unchanged across unrelated substrates. Prime promotion would duplicate Classification, Measurement, Difference, and Threshold while importing contested psychological content into an allegedly universal node.

Instantiates / Related Primes

Classification is the minimal proposed parent. E–S theory presupposes a rule that maps standardized relative scores to named profile classes. Classification can occur without empathising, systemising, psychometrics, sex-coded comparisons, or autism, so the direction is strict.

Measurement is a constitutive related prime: each dimension must be operationalized, scored, standardized, and evaluated for validity. It is not proposed as an additional parent because particular EQ/SQ instruments are replaceable and the minimal DAG should state the narrowest stable genus rather than every dependency.

Difference and Threshold describe the D = S - E transformation and category cut points. Theory of Mind overlaps the mental-state-inference portion of empathising but not affective response, systemising, or classification. Stereotyping identifies a misuse route when group averages are converted into deterministic judgments about individuals; it does not cover the theory's measurement-and-test structure. Systems Thinking is a lexical neighbor, not an ancestor.

Relationships to Other Abstractions

Current abstraction Empathising–Systemising Theory Domain-specific

Parents (1) — more general patterns this builds on

• Empathising–Systemising Theory presupposes Classification Prime

  Classification is the minimal proposed parent.

Hierarchy path (1) — routes to 1 parentless root

• Empathising–Systemising Theory → Classification

Neighborhood in Abstraction Space

Empathising–Systemising Theory sits in a sparse region of the domain-specific corpus (87^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

• Identification (Psychology) — 0.83
• Personality Change — 0.80
• Systemic Hypothesising — 0.79
• Authoritarian Personality — 0.79
• Subpersonality — 0.79

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With

• Empathy: broader cognitive, affective, embodied, and relational phenomena; E–S operationalizes a particular compound construct.
• Cognitive versus affective empathy: analytically separable components whose evidence should not be collapsed into one deficit claim.
• Theory of Mind: mental-state attribution without the entire E–S pair and profile taxonomy.
• Systems Thinking: a reasoning practice concerned with interdependence and feedback, not the SQ-defined disposition.
• System 1/System 2: a dual-process account unrelated to the E–S use of systemising.
• EQ, SQ, SQ-R, and AQ: instruments and scores, not the theory; AQ is not a diagnosis.
• Autism diagnosis: a clinical assessment cannot be inferred from Type S or Extreme S.
• Extreme Male Brain theory: an influential, stronger autism-related extension rather than an exact alias for every E–S use.
• Hyper-systemising: a proposed mechanism or degree claim, not the complete two-dimension theory.
• Sex, gender identity, and gender role: distinct variables that a study must measure and name accurately.
• Gender essentialism: an ideological or causal conclusion not entailed by overlapping group-level distributions.
• Big Five agreeableness: a strong empirical neighbor of EQ in some studies, but not the E–S classification operation.^[6]
• Technical or STEM ability: an SQ score or profile is not a credential, achievement test, or destiny.

References

[1] Baron-Cohen, S. (2002). “The extreme male brain theory of autism.” Trends in Cognitive Sciences, 6(6), 248–254. doi:10.1016/S1364-6613(02)01904-6. registry ↩a ↩b

[2] Baron-Cohen, S., Richler, J., Bisarya, D., Gurunathan, N., & Wheelwright, S. (2003). “The Systemizing Quotient: An investigation of adults with Asperger syndrome or high-functioning autism, and normal sex differences.” Philosophical Transactions of the Royal Society B, 358, 361–374. doi:10.1098/rstb.2002.1206. registry ↩a ↩b ↩c

[3] Greenberg, D. M., Warrier, V., Allison, C., & Baron-Cohen, S. (2018). “Testing the Empathizing–Systemizing theory of sex differences and the Extreme Male Brain theory of autism in half a million people.” Proceedings of the National Academy of Sciences, 115, 12152–12157. doi:10.1073/pnas.1811032115. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[4] Baron-Cohen, S., & Wheelwright, S. (2004). “The Empathy Quotient: An investigation of adults with Asperger syndrome or high functioning autism, and normal sex differences.” Journal of Autism and Developmental Disorders, 34, 163–175. doi:10.1023/B:JADD.0000022607.19833.00. registry ↩a ↩b

[5] Cusson, S., et al. (2025). “Quantifying empathy differences in autism: A systematic review and meta-analysis.” Clinical Psychology Review, 120, 102623. doi:10.1016/j.cpr.2025.102623. registry ↩a ↩b ↩c

[6] Nettle, D. (2007). “Empathizing and systemizing: What are they, and what do they contribute to our understanding of psychological sex differences?” British Journal of Psychology, 98, 237–255. doi:10.1348/000712606X117612. registry ↩a ↩b ↩c

[7] Groen, Y., et al. (2015). “The Empathy and Systemizing Quotient: The psychometric properties of the Dutch version and a review of the cross-cultural stability of the factor structure and sex differences.” Journal of Autism and Developmental Disorders, 45, 2848–2864. doi:10.1007/s10803-015-2448-z. registry ↩a ↩b

[8] Milton, D. E. M. (2012). “On the ontological status of autism: The ‘double empathy problem’.” Disability & Society, 27, 883–887. doi:10.1080/09687599.2012.710008. registry ↩a ↩b

[9] Harrison, J. L., Brownlow, C., Ireland, M. J., & Piovesana, A. (2022). “Empathy measurement in autistic and nonautistic adults: A COSMIN systematic literature review.” Assessment, 29(2), 332–350. doi:10.1177/1073191120964564. registry ↩

[10] Lawrence, E. J., Shaw, P., Baker, D., Baron-Cohen, S., & David, A. S. (2004). “Measuring empathy: reliability and validity of the Empathy Quotient.” Psychological Medicine, 34, 911–919. doi:10.1017/S0033291703001624. registry

[11] Lai, M.-C., Lombardo, M. V., Auyeung, B., Chakrabarti, B., & Baron-Cohen, S. (2015). “Sex/gender differences and autism: Setting the scene for future research.” Journal of the American Academy of Child & Adolescent Psychiatry, 54, 11–24. doi:10.1016/j.jaac.2014.10.003. registry