OLM 4: Correlation vs Causation

Causal vs Correlational: Part 1

Key distinction:
- Correlational study: Measures how two variables are related; no manipulation of variables; cannot determine causation—only association.
- Causal (experimental) study: Researcher manipulates an independent variable; measures its effect on a dependent variable; uses random assignment to control for confounds; causality can be inferred if the design is strong.
Example illustrating correlation without causation:
- Childhood maltreatment & adult offending (Widom, 1989): positive correlation found, but other factors (e.g., poverty, mental illness) not ruled out.
- Citation: Widom, C. S. (1989). The cycle of violence. Science, 244(4901), 160–166.

What is a correlational study?

Measures the relationship between two variables without any manipulation.
Cannot determine directionality or causation; only association.
Example: Childhood maltreatment and adult offending showing a correlation, but not proving maltreatment causes offending due to potential confounds.
Real-world example cited: Widom (1989).

What is a causal study?

Researchers manipulate an independent variable (IV).
Observe effects on a dependent variable (DV).
Use random assignment to control for confounds and reduce selection bias.
Causality can be inferred if the design is strong (e.g., randomized controlled trials).
Real-world example cited: Anger management program to reduce reoffending (Howells et al., 2005): Randomised control trial comparing intervention group to no-treatment group.
Citation: Howells, K., Day, A., & Thomas-Peter, B. (2004). Journal of Forensic Psychiatry & Psychology, 15(2), 359–385.

Correlation ≠ Causation

Core idea: Co-occurrence does not establish one variable causes the other.
Example: Low empathy scores correlate with violent offending.
- Question prompts:
- Does low empathy cause violence?
- Or does exposure to violence reduce empathy?
- Or are both caused by a third factor (e.g., early trauma)?
Term to know: Spurious correlation = when two variables appear related due to a third variable.
Implication: Correlations can guide hypotheses, but require experimental tests to establish causality.

Strengths & Weaknesses – Correlational Studies

Strengths:
- Ethical and non-invasive
- Allows exploration of naturally occurring variables
- Useful for hypothesis generation
Weaknesses:
- No control over confounding variables
- Cannot infer cause and effect
Example: fMRI studies showing correlation between brain activity differences and psychopathy (e.g., Raine, 1993) — reduced prefrontal activity in violent offenders.

Strengths & Weaknesses – Experimental Studies

Strengths:
- High internal validity
- Can test cause-and-effect relationships
Weaknesses:
- Ethical limitations
- May lack ecological validity
- Artificial settings (e.g., labs or prisons)
Example: It is often unethical to experimentally induce trauma or antisocial behaviour to test long-term outcomes.

When to use each design?

Design type → Best used when:
- Correlational: Ethical or practical limits prevent manipulation; good for exploring relationships.
- Experimental: To test if X causes Y and can manipulate conditions.
Tip: Ask yourself — Am I just observing, or am I testing a change?

Summary

Correlational studies identify associations, but not direction or causality.
Experimental designs allow stronger causal inferences, but require careful control.
Both designs play key roles in forensic psychology.

Causal vs Correlational: Part 2

Why comparing the two designs?

UNDERSTAND HOW METHODOLOGY INFLUENCES FINDINGS
IDENTIFY STRENGTHS AND LIMITATIONS OF EACH APPROACH
LEARN HOW RESEARCH QUESTIONS SHAPE DESIGN CHOICES

Study 1 – Correlational Design

Sample: $n = 269$ male offenders, ages $18 ext{-}25$
Focus: Neuropsychological performance (e.g., response inhibition) correlated with aggression levels
Key Finding: Poor response inhibition was significantly associated with higher aggression scores—but correlational, no manipulation.
Conclusion: Cognitive deficits relate to aggression, but causal direction remains unclear.
Citation: Wallinius, M., Nordholm, J., Wagnström, F., & Billstedt, E. (2019). Cognitive functioning and aggressive antisocial behaviors in young violent offenders. Psychiatry research, 272, 572–580. https://doi.org/10.1016/j.psychres.2018.12.140

Study 1 – Correlational Design Strengths

Large, real-world forensic sample
Objective neuropsychological measures
Strong for establishing associations

Study 1 – Correlational Design Limitations

Correlational = no causal inference
Possible third variables (e.g., trauma)
Cross-sectional design

Study 2 – Experimental Design

Sample: $n = 84$ young male offenders, ages $16 ext{-}18$
Design: Randomized controlled trial (RCT) with ABM vs placebo vs waitlist
Intervention: 4-week attention bias training redirecting focus from hostile cues
Key Finding: Offenders with high initial hostile bias showed significant reductions in aggression post-ABM.
Conclusion: Supports a causal role for attentional biases in aggression.
Citation: Zhao, Z., Yu, X., Ren, Z., Zhang, L., & Li, X. (2022). The remediating effect of Attention Bias Modification on aggression in young offenders with antisocial tendency: A randomized controlled trial. Journal of behavior therapy and experimental psychiatry, 75, 101711. https://doi.org/10.1016/j.jbtep.2021.101711

Study 2 – Experimental Design Strengths

Random assignment → stronger inference of causality
Direct intervention targeting mechanism
Controlled design with placebo comparison

Study 2 – Experimental Design Limitations

Small sample limits generalizability
Only male participants
Short intervention and follow-up period

Comparison of the Methodologies

Study 1 (Correlational)
- Design: Cross-sectional
- Sample Size: $n = 269$
- Manipulation of Variable: No
- Causal Claim: No
- Measure Type: Neuropsychological tests
- Ecological Validity: Moderate
Study 2 (Experimental)
- Design: RCT
- Sample Size: $n = 84$
- Manipulation of Variable: Yes
- Causal Claim: Yes (conditional)
- Measure Type: Behavioral intervention
- Ecological Validity: Moderate–low

Bringing the findings from both together

Cognitive deficits (like poor inhibition) are linked to aggression → targetable risk markers
Attention training can reduce aggression — suggests attentional bias is not just correlated but can be changed
Combined insight: Issues identified in correlational designs may become effective intervention targets in experimental designs

Conclusion

Correlational studies highlight risk mechanisms.
Experimental interventions test whether altering those mechanisms reduces aggression.
Together, they offer a fuller picture—from explanation to action

Notes on examples and context

Classic correlational example: Widom (1989) linking childhood maltreatment to later offending, highlighting the need to account for confounds such as poverty or mental illness.
Experimental ethics: It is often unethical or impractical to manipulate trauma exposure or reproduce antisocial behavior in participants to study long-term outcomes; hence, reliance on ethically designed interventions and simulations.
Practical relevance: Forensic psychology uses both designs to identify risk markers (correlational) and to test interventions that target mechanistic processes (experimental).
Real-world relevance: Findings support the development of interventions like attention bias modification (ABM) to reduce aggression among youth offenders.

Key definitions recap

Correlational study: Measures association between two variables without manipulation; cannot infer causality.
Causal (experimental) study: Manipulates IV, uses randomization; can support causal inferences.
Spurious correlation: A correlation that is due to a third, unmeasured variable.
Randomized Controlled Trial (RCT): Gold standard for testing causal effects by random assignment to conditions.
Attention Bias Modification (ABM): Intervention designed to redirect attention away from hostile cues, used here as a causal test of attentional bias in aggression.