Concepts · Dynamic Systems and Integrated Change

Reciprocal Causation

Behaviour is not always the final outcome of a causal process. What a person does can change the conditions, responses and information that shape what happens next. Reciprocal causation describes this bidirectional influence across time—without implying equal power, simultaneity or a feedback loop.

By Yona Ole Lobulu ·

Concept8 min readFoundationalD12.2

Topic
Dynamic Systems and Integrated Change
Read first
One piece should be read before this one
Reading time
About 8 minutes of reading
Difficulty
Late reading: this page sits at the far end of the Library, after many other pieces

The question

What changes when causes and outcomes influence one another rather than operating only in one direction?

Definition

Reciprocal causation occurs when one process causally influences another and the resulting change subsequently affects the original process or its causal conditions, producing causal influence in both directions across time.

Behaviour is often treated as the endpoint of explanation. An environment shapes what a person does; a belief influences an action; another person's response changes behaviour. But sometimes the causal process does not stop there.

A person's behaviour can alter the social response they receive. An action can change what information becomes available. Repeated behaviour can modify aspects of the environment encountered later. What begins as an outcome can therefore become part of the causal conditions shaping what happens next.

Behaviour can be an outcome at one moment and part of the causal conditions for what happens next.

That is the central idea of reciprocal causation.

Causes and outcomes do not have permanent roles

A simple causal relation can be represented as A₁ → B₂. A condition at one time contributes causally to a later outcome.

Reciprocal causation adds another direction across time: A₁ → B₂ → A₃. A influences B. The change in B then affects the original process, or the conditions from which it arose, at a later point.

The essential shift is temporal. A variable that functions as an outcome in one causal relation can become a cause in another relation later.

This does not make the categories cause and outcome meaningless. They remain meaningful relative to a particular causal relation and time point. What changes is the assumption that one variable must remain permanently upstream while another remains permanently downstream.

Human Behaviour Has Multiple Causes established that behaviour can have multiple causal contributors. Reciprocal causation adds that causes, outcomes and surrounding conditions can alter one another across time.

Not every causal relationship is reciprocal, and not every behaviour meaningfully changes the conditions that produced it. Reciprocal causation describes a particular kind of causal architecture, not a universal property of human behaviour.

Reciprocal causation preserves direction

Reciprocity is often compressed into a double arrow, A ↔ B.

That notation can be useful, but it can also hide the temporal structure. The more informative representation is A₁ → B₂, followed later by B₂ → A₃.

Each relation still has a direction. A changes B; B later contributes to changing A or its causal conditions.

Reciprocal causation therefore adds a second causal direction. It does not remove causal direction.

Nor does reciprocity require both effects to occur simultaneously. One may unfold quickly while the reverse influence develops later. What matters is causal influence in both directions across time.

Reciprocal does not mean equal

Two causal directions can exist without being equally strong.

Consider a workplace. Scheduling rules, staffing levels and organizational policies may strongly constrain an employee's behaviour. The employee's behaviour may also affect later workplace conditions by changing what managers observe, how tasks are allocated or how colleagues respond.

Those effects need not be comparable in magnitude.

The institution may exert far greater causal influence on the individual than the individual exerts on the institution. Both directions can still be causally relevant.

The same principle applies in interpersonal relationships. One person's behaviour may affect another person's later response even when differences in authority, resources or dependence make the causal relationship highly asymmetric.

Bidirectional causation does not imply equal causal power.

Reciprocal influence also does not, by itself, establish responsibility, blame or equal agency. Those are separate questions.

Behaviour can change its future conditions

The idea becomes clearer when the causal sequence is made concrete.

An environment can shape what actions are available, costly or likely. Behaviour can then alter some part of the environment encountered later.

Suppose a social setting initially discourages participation. A person participates only minimally. Other people receive less information from them and later involve them less in discussion. The social conditions encountered afterward are now partly different from those encountered at the beginning.

The original environment influenced behaviour. The behaviour then changed part of the later social environment, which can influence subsequent behaviour in return.

This is the basic causal architecture developed more fully in Person–Environment Transactions.

A similar structure can occur across cognition and behaviour. An expectation may influence an action; the action produces consequences and information; those consequences may later alter the expectation. In simplified form: expectation₁ → action₂ → consequence₃ → later expectation₄.

This does not imply that every consequence changes a belief or that one theory of belief updating explains the process. It shows only that a state which initially helped cause behaviour can later be altered by the consequences of that behaviour.

Interpersonal processes provide another case. Another person's response can influence how someone behaves. That behaviour changes what the other person encounters and may alter how they respond later.

In each example, behaviour is not merely a passive endpoint. It changes some part of the causal landscape that comes next.

Influencing an environment is not the same as controlling it

Reciprocal causation can easily be overstated.

If behaviour changes later environmental conditions, it may sound as though people simply create the environments that shape them. That conclusion does not follow.

Many environments contain conditions that individuals did not choose and have little power to alter. Physical constraints, laws, institutional structures, unequal resources and other people's decisions can strongly limit what behaviour can change.

Behaviour may affect only a small portion of the relevant environment. Some structural conditions may resist individual influence almost entirely, while meaningful change may require coordinated or collective action.

Influencing an environment is not the same as controlling it.

Institutions can strongly shape individual behaviour while repeated individual and collective practices may also contribute to institutional persistence or change. Reciprocity therefore remains compatible with strong structural constraint and large inequalities in causal power.

Two-way prediction is not two-way causation

Reciprocal causal hypotheses are easier to formulate than to establish empirically.

Suppose a longitudinal study repeatedly measures A and B. Earlier A predicts later B, and earlier B predicts later A. That pattern is compatible with reciprocal causation. It does not prove it.

A third factor may influence both variables. Stable differences between people may generate an apparent reciprocal pattern. Measurement problems, poorly chosen time intervals or assumptions built into the statistical model can also affect the result.

Traditional cross-lagged models have often been used to study reciprocal relations, but methodological work has shown that they can combine stable between-person differences with within-person temporal variation in ways that complicate interpretation.

This distinction matters. A pattern showing that people who are generally high on A are also generally high on B is not the same as showing that, when A changes within a particular person, B subsequently changes.

More refined longitudinal models can separate some of these sources of variation. That can improve the match between the statistical model and the causal question. It still does not make the resulting coefficient causal by itself.

Causal interpretation also depends on alternative explanations, measurement, temporal order, model assumptions and whether the design can actually identify the causal effect being claimed.

This is the same evidential discipline established in Correlation, Prediction, Causation and Mechanism, now applied in both directions.

Two-way prediction is not automatically evidence of two-way causation.

Temporal order matters, but it is not enough

Time remains essential.

If A is proposed to cause B, the relevant change in A must precede the change in B attributed to it. If B later causes A, that second relation also requires an appropriate temporal ordering.

But precedence alone does not establish causation.

Earlier A may predict later B because both were influenced by an unmeasured cause. The interval between measurements may also be poorly matched to the process. A relation unfolding over hours may be difficult to detect with observations separated by years.

Reciprocal effects therefore need not unfold at the same rate, and detecting them depends partly on measuring the process at an appropriate timescale.

Temporal order is part of causal evidence, not a substitute for it.

Stronger reciprocal causal claims require evidence capable of supporting both directional effects while addressing plausible alternatives. No single pattern of longitudinal association guarantees that.

Reciprocal causation is not yet a feedback loop

The distinction between reciprocal causation and feedback is crucial.

A reciprocal causal sequence can be represented as A₁ → B₂ → A₃. This establishes influence in both directions across time. It does not tell us whether the relation continues.

A recurrent structure would look more like A₁ → B₂ → A₃ → B₄ → A₅ …

Once the consequences repeatedly re-enter the process, new questions arise about whether the recurring effects sustain, amplify or counteract change. Those are questions about feedback.

Reciprocal causal relations can form part of feedback processes, but reciprocity alone does not establish a feedback loop.

That distinction preserves the boundary between this concept and Feedback Loops Sustain Patterns.

From multiple causes to dynamic causes

Multiple causation established that human behaviour can depend on several causal contributions. Reciprocal causation changes the picture again.

Behaviour is not always the final box at the end of a causal diagram. An outcome can alter conditions that shape what happens later. Environments can affect behaviour while behaviour changes aspects of later environments. Expectations can influence actions while the consequences of those actions alter later expectations. Relationships can shape people while behaviour changes later relationship conditions.

These relations remain directional. They need not be simultaneous or equally strong, and reciprocal prediction does not by itself establish reciprocal causation.

The central shift is that causal roles can change across time.

A process can help produce an outcome and later be altered by that outcome or its consequences, creating causal influence in both directions without dissolving causal order.

That moves the Library from multiple causation toward dynamic causal structure.

The next question is what happens when reciprocal causal relations do not occur only once, but continue to re-enter the system. That is the problem of Feedback Loops Sustain Patterns.

Sources and research record5 sources, with findings, strengths and limitations as entered

References

5 sources this piece rests on, as entered in the Library.

  1. Bell, R. Q. (1968) A Reinterpretation of the Direction of Effects in Studies of Socialization

    Theoretical review · Psychological Review, 75(2) · 81–95

    The historical challenge to one-way socialization models: children's behaviour can influence adults and the social conditions encountered later.

    doi:10.1037/h0025583

  2. Bandura, A. (1989) Human Agency in Social Cognitive Theory

    Theoretical article · American Psychologist, 44(9) · 1175–1184

    Triadic reciprocal causation among personal factors, behaviour and environment, with behaviour as an active causal participant rather than only an endpoint. An influential framework, not the Library's definition.

    doi:10.1037/0003-066X.44.9.1175

  3. Sameroff, A. (2009) The Transactional Model of Development: How Children and Contexts Shape Each Other

    Edited volume · American Psychological Association

    Continuing person–context transactions and reciprocal developmental influence unfolding over time.

    doi:10.1037/11877-000

  4. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

    Methodological article · Psychological Methods, 20(1) · 102–116

    Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

    doi:10.1037/a0038889

  5. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

    Methodological article · Psychological Methods, 24(5) · 637–657

    Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

    doi:10.1037/met0000210

Further reading

Behind this page

The claims this concept makes, the evidence behind them, and the limits it accepts.

Evidence status

Established

Well supported by a substantial, converging empirical literature.

Claims

  1. An outcome can become part of the causal conditions for what happens later

    Established

    What this does not assert: A property some causal relations have, not a universal feature of behaviour.

    1. Bell, R. Q. (1968) A Reinterpretation of the Direction of Effects in Studies of Socialization

      Theoretical review · Psychological Review, 75(2) · 81–95

      The historical challenge to one-way socialization models: children's behaviour can influence adults and the social conditions encountered later.

      doi:10.1037/h0025583

    2. Bandura, A. (1989) Human Agency in Social Cognitive Theory

      Theoretical article · American Psychologist, 44(9) · 1175–1184

      Triadic reciprocal causation among personal factors, behaviour and environment, with behaviour as an active causal participant rather than only an endpoint. An influential framework, not the Library's definition.

      doi:10.1037/0003-066X.44.9.1175

  2. Traditional cross-lagged models can combine stable between-person differences with within-person temporal variation

    Established

    What this does not assert: A modelling limitation, not evidence that reciprocal effects are absent.

    1. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

      Methodological article · Psychological Methods, 20(1) · 102–116

      Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

      doi:10.1037/a0038889

  3. Between-person associations and within-person temporal processes answer different questions

    Established

    What this does not assert: Neither answer substitutes for the other.

    1. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

      Methodological article · Psychological Methods, 20(1) · 102–116

      Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

      doi:10.1037/a0038889

  4. More refined longitudinal models can improve process identification without establishing causality

    Established

    What this does not assert: A better match to the causal question is not a causal warrant.

    1. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

      Methodological article · Psychological Methods, 24(5) · 637–657

      Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

      doi:10.1037/met0000210

  5. Temporal precedence contributes to causal inference but is insufficient by itself

    Established

    What this does not assert: An unmeasured common cause can produce the same ordering.

    1. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

      Methodological article · Psychological Methods, 24(5) · 637–657

      Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

      doi:10.1037/met0000210

  6. Measurement interval and timescale can materially affect estimated reciprocal relations

    Established

    What this does not assert: A process unfolding over hours may be invisible across years.

    1. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

      Methodological article · Psychological Methods, 24(5) · 637–657

      Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

      doi:10.1037/met0000210

  7. Cognition and behaviour can participate in reciprocal causal relations

    High confidence

    What this does not assert: An example of the architecture, not a theory of belief updating.

    1. Bandura, A. (1989) Human Agency in Social Cognitive Theory

      Theoretical article · American Psychologist, 44(9) · 1175–1184

      Triadic reciprocal causation among personal factors, behaviour and environment, with behaviour as an active causal participant rather than only an endpoint. An influential framework, not the Library's definition.

      doi:10.1037/0003-066X.44.9.1175

  8. People and environments can participate in transactional processes

    High confidence

    What this does not assert: Framework-sensitive; the full transactional theory belongs downstream.

    1. Sameroff, A. (2009) The Transactional Model of Development: How Children and Contexts Shape Each Other

      Edited volume · American Psychological Association

      Continuing person–context transactions and reciprocal developmental influence unfolding over time.

      doi:10.1037/11877-000

  9. Interpersonal responses can influence later behaviour while behaviour alters later interpersonal conditions

    High confidence

    What this does not assert: Holds for some relationships, not all.

    1. Bell, R. Q. (1968) A Reinterpretation of the Direction of Effects in Studies of Socialization

      Theoretical review · Psychological Review, 75(2) · 81–95

      The historical challenge to one-way socialization models: children's behaviour can influence adults and the social conditions encountered later.

      doi:10.1037/h0025583

  10. Repeated individual and collective practices can sometimes contribute to institutional persistence or change

    Canonical inference

    What this does not assert: Compatible with individual influence being far smaller than institutional influence.

  11. Influencing an environment is not the same as controlling it

    Canonical inference

    What this does not assert: Much of the relevant environment may resist individual influence entirely.

  12. Behaviour can function as a causal outcome at one stage and as a causal influence at a later stage

    Established

    What this does not assert: Causal roles are relative to a relation and a time point, not permanent properties.

    1. Bandura, A. (1989) Human Agency in Social Cognitive Theory

      Theoretical article · American Psychologist, 44(9) · 1175–1184

      Triadic reciprocal causation among personal factors, behaviour and environment, with behaviour as an active causal participant rather than only an endpoint. An influential framework, not the Library's definition.

      doi:10.1037/0003-066X.44.9.1175

  13. Reciprocal influence does not by itself establish responsibility or blame

    Canonical inference

    What this does not assert: A separate normative question from the causal one.

  14. Reciprocal causation is not equivalent to feedback

    Canonical inference

    What this does not assert: Reciprocity is a causal relation; feedback is a recurrent causal dynamic.

  15. Reciprocal causal relations can participate in later feedback processes

    Canonical inference

    What this does not assert: Participation is possible, not automatic; recurrence must be shown separately.

  16. Reciprocal causation preserves causal direction and temporal order

    Established

    What this does not assert: A second direction is added; no direction is dissolved.

    1. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

      Methodological article · Psychological Methods, 20(1) · 102–116

      Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

      doi:10.1037/a0038889

  17. Reciprocal effects need not be simultaneous

    Established

    What this does not assert: One direction may unfold quickly while the reverse influence develops later.

    1. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

      Methodological article · Psychological Methods, 24(5) · 637–657

      Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

      doi:10.1037/met0000210

  18. Reciprocal effects need not be equally strong

    Established

    What this does not assert: Bidirectional causation does not imply equal causal power.

    1. Bell, R. Q. (1968) A Reinterpretation of the Direction of Effects in Studies of Socialization

      Theoretical review · Psychological Review, 75(2) · 81–95

      The historical challenge to one-way socialization models: children's behaviour can influence adults and the social conditions encountered later.

      doi:10.1037/h0025583

  19. Person–environment and interpersonal processes can be studied transactionally rather than as purely one-way influence

    Established

    What this does not assert: Transactional method does not by itself settle the size or direction of any effect.

    1. Sameroff, A. (2009) The Transactional Model of Development: How Children and Contexts Shape Each Other

      Edited volume · American Psychological Association

      Continuing person–context transactions and reciprocal developmental influence unfolding over time.

      doi:10.1037/11877-000

  20. Reciprocal causal architecture is compatible with strong structural constraint and asymmetric causal power

    Established

    What this does not assert: Influence in both directions is not symmetry of power.

    1. Bell, R. Q. (1968) A Reinterpretation of the Direction of Effects in Studies of Socialization

      Theoretical review · Psychological Review, 75(2) · 81–95

      The historical challenge to one-way socialization models: children's behaviour can influence adults and the social conditions encountered later.

      doi:10.1037/h0025583

  21. Cross-sectional association cannot establish reciprocal causation

    Established

    What this does not assert: Association at one time point carries no temporal structure at all.

    1. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

      Methodological article · Psychological Methods, 20(1) · 102–116

      Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

      doi:10.1037/a0038889

  22. Bidirectional longitudinal prediction does not by itself establish reciprocal causation

    Established

    What this does not assert: Compatible with reciprocity; not proof of it.

    1. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

      Methodological article · Psychological Methods, 20(1) · 102–116

      Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

      doi:10.1037/a0038889

    2. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

      Methodological article · Psychological Methods, 24(5) · 637–657

      Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

      doi:10.1037/met0000210

Sources

  1. Bell, R. Q. (1968) A Reinterpretation of the Direction of Effects in Studies of Socialization

    Theoretical review · Psychological Review, 75(2) · 81–95

    The historical challenge to one-way socialization models: children's behaviour can influence adults and the social conditions encountered later.

    doi:10.1037/h0025583

  2. Bandura, A. (1989) Human Agency in Social Cognitive Theory

    Theoretical article · American Psychologist, 44(9) · 1175–1184

    Triadic reciprocal causation among personal factors, behaviour and environment, with behaviour as an active causal participant rather than only an endpoint. An influential framework, not the Library's definition.

    doi:10.1037/0003-066X.44.9.1175

  3. Sameroff, A. (2009) The Transactional Model of Development: How Children and Contexts Shape Each Other

    Edited volume · American Psychological Association

    Continuing person–context transactions and reciprocal developmental influence unfolding over time.

    doi:10.1037/11877-000

  4. Hamaker, E. L., Kuiper, R. M., Grasman, R. P. P. P. (2015) A Critique of the Cross-Lagged Panel Model

    Methodological article · Psychological Methods, 20(1) · 102–116

    Conventional cross-lagged models can combine stable between-person differences with within-person temporal variation, complicating reciprocal-effects interpretation.

    doi:10.1037/a0038889

  5. Usami, S., Murayama, K., Hamaker, E. L. (2019) A Unified Framework of Longitudinal Models to Examine Reciprocal Relations

    Methodological article · Psychological Methods, 24(5) · 637–657

    Temporal modelling of reciprocal relations, lag selection, and the limits of longitudinal prediction as causal identification.

    doi:10.1037/met0000210

What this opens up

What becomes readable once you have this.

Where to go from here

Next published piece

Feedback Loops Sustain Patterns

Feedback loops allow the consequences of a process to return and influence what happens next. Reinforcing loops can amplify or reproduce a pattern, while balancing loops can counteract change and keep a system within a familiar range. Patterns therefore persist not only because an original cause remains present, but because repeated outcomes can recreate some of the conditions that produced them.

Continue through the Library →See where this sits in the graph →

Back to the Library →