Introduction to Clinical Research
Topic overview
Clinical researchers aim to uncover nomothetic principles of abnormal psychological functioning using the scientific method.
They seek general laws or principles that apply across individuals (nomothetic understanding) rather than focusing on a single case (idiographic understanding).
Key goals:
Describe relationships between variables.
Determine whether changes in one variable relate to changes in another.
Test predictions and hypotheses about abnormality across many individuals.
Important terms:
Variable: any characteristic or event that can vary across time, place, or person. Examples include childhood upsets, present life experiences, moods, social functioning, and responses to treatment.
Independent variable (IV): the manipulated factor in an experiment.
Dependent variable (DV): the factor observed and measured for change.
Sample vs population: the subset studied vs. the larger group of interest.
Internal validity: accuracy in attributing observed effects to the manipulated variable.
External validity: generalizability of findings to other people and settings.
Three main methods of investigation used by clinical researchers (each best suited to different questions):
Case study
Correlational method
Experimental method
Core reason for using multiple methods: to form and test hypotheses and to draw broad conclusions about why certain variables are related.
Key caveat: scientific progress often requires testing ideas rather than relying on conventional wisdom alone (e.g., lobotomies and psychoanalytic ideas once thought true).
Ethical and practical constraints shape how researchers study abnormal psychology (e.g., measurement of elusive constructs, cultural variability, rights of participants).
The case study
Definition: a detailed description of a person’s life and psychological problems, including history, present circumstances, symptoms, possible causes, and sometimes treatment.
Uses and value:
Source of new ideas about behavior.
May open pathways for discoveries and generate hypotheses.
May provide tentative support for a theory or show the value of new therapies.
Useful for studying unusual problems that do not occur frequently enough for large samples.
Famous example: Freud’s Little Hans case study (1909) describing a 4-year-old boy with a fear of horses; includes material from Hans’s father and Freud’s interpretations, leading to insights about repression, castration anxiety, and the family dynamics surrounding a child’s phobia.
Other influential case studies: Oliver Sacks’ The Man Who Mistook His Wife for a Hat (1985) illustrating various neurological cases and broad psychological processes.
How case studies can play nomothetic roles beyond the individual case and influence theories and therapies.
Limitations of case studies:
Bias: cases are reported by therapists who may have a stake in outcomes; selection bias in what gets reported.
Subjectivity: evidence is often interpretive and depends on the therapist’s or client’s reports.
Internal validity: case studies typically have low internal validity because they do not control for confounding variables.
External validity: limited generalizability to other individuals or populations.
Notable limitations illustrated by famous cases: whether a single case proves a general rule; extrapolating from one individual to many.
Notable discussion points from the case-study literature:
Case studies can inform theory and therapy, but they do not establish causality.
They may contribute to understanding patterns across families or cultures (e.g., dissociative identity disorder cases like the “Three Faces of Eve”).
The correlational method
Definition: a research procedure used to determine the degree to which two or more variables vary together.
Key ideas:
Correlation describes a relationship or association, not causation.
If two variables vary together, they may be related; this can be positive, negative, or non-existent.
Describing a correlation:
Scatterplots: each point represents a participant’s scores on two variables.
Line of best fit: the straight line that minimizes distance between data points and the line, illustrating the strength and direction of the relationship.
Direction of correlation:
Positive correlation: as one variable increases, the other tends to increase (line slopes upward).
Negative correlation: as one variable increases, the other tends to decrease (line slopes downward).
No correlation: no consistent relationship (line is flat).
Magnitude (strength) of correlation: described by the correlation coefficient r \in [-1, 1]. The closer |r| is to 1, the stronger the relationship; the closer to 0, the weaker the relationship.
Examples:
Positive: life stress and depression tend to rise together (e.g., a strong positive trend; a line of best fit with a steep slope).
Negative: depression and activity level tend to move in opposite directions (higher depression associated with fewer activities; negative slope).
No correlation: intelligence and depression may show near-zero correlation in some samples.
Magnitude examples from the transcript:
A strong positive correlation example had a line of best fit with a steep slope and data points close to the line (e.g., r around +0.53 in some long-running studies).
A moderately positive correlation example had a less steep line (e.g., r around +0.28).
A strong negative correlation example and a near-zero correlation example were also described.
Describing sample validity:
Sample should be representative of the population to generalize findings (external validity).
If a sample is not representative (e.g., only children), generalizability to adults may be compromised.
Statistical significance in correlational findings:
Researchers use probability to determine whether a correlation is unlikely due to chance.
Convention: if there is less than a 5% probability that the observed correlation occurred by chance, the result is considered statistically significant.
Notation: significance is often expressed as p < 0.05.
Practical interpretation:
Larger magnitude and larger sample sizes increase confidence that a correlation reflects a real association in the population.
Limitations:
Correlation does not imply causation; third variables or bidirectional influences may explain the association.
The method is strong for identifying relationships but weak for establishing causal directions.
The experimental method
Definition: a research procedure in which a variable is manipulated and the manipulation’s effect on another variable is observed.
Core concepts:
Independent variable (IV): the manipulated factor believed to cause an effect.
Dependent variable (DV): the outcome measured to assess the effect of the IV.
Random assignment: participants are assigned to groups in a way that each participant has an equal chance of being placed in any group, helping to control preexisting differences.
Control group: a group not exposed to the IV, used for comparison to determine the IV’s effect.
Experimental group: the group exposed to the IV.
Mask design (blinding): procedures to prevent participants or researchers from knowing group assignments to reduce bias.
How experiments test causal questions:
Example question: Does a particular therapy relieve symptoms of a disorder?
Therapy is the IV; psychological improvement is the DV.
Statistical significance and clinical significance:
Statistical significance: when the observed difference between groups is unlikely to have occurred by chance, typically p < 0.05.
Clinical significance: whether the amount of improvement is meaningful in the participant’s life, beyond statistical metrics.
Random assignment and confounds:
Random assignment helps ensure comparable groups and reduces preexisting differences as potential confounds.
Confounds: other variables that might influence the DV besides the IV (e.g., office location, soothing music, participant motivation).
To guard against confounds, experiments typically use a control group, random assignment, and mask design.
Practical considerations:
In clinical research, there are ethical and practical limits on manipulations; designs may be less than ideal and incorporate quasi-experimental elements.
Statistical analyses assess whether observed group differences are likely due to the IV rather than chance.
Key distinctions:
Statistical significance vs. clinical significance: a result can be statistically significant but not necessarily meaningful in real-world life quality improvements.
Deterministic vs probabilistic conclusions:
When true causation cannot be separated from other potential causes, experiments provide limited information.
Alternative (quasi-experimental) research designs
When pure experiments are not feasible or ethical, researchers use quasi-experimental designs that mix elements of experimental and correlational methods:
Matched design
Natural experiment
Analog experiment
Single-case experiment
Longitudinal study
Epidemiological study
Matched design:
Researchers compare groups that are already existing in the world (e.g., abused vs. non-abused children) and match participants on key characteristics (age, sex, race, socioeconomic status, etc.) to reduce confounds.
Natural experiments:
Nature manipulates the IV; researchers observe effects. Examples include studying effects of floods, earthquakes, plane crashes, or fires.
Useful for studying unusual or unpredictable events, but generalization can be limited because events are not controlled.
Classic example discussions include tsunamis (Sumatra 2004) and other disasters (Haiti 2010, Japan 2011, Sandy 2012, various California wildfires 2018–2019).
Analog experiments:
Researchers induce abnormal-like behavior in laboratory participants (humans or animals) and study the outcomes to shed light on real-world conditions.
Often used to explore causes of human depression (learned helplessness paradigm).
Major caveat: laboratory phenomena may not perfectly map onto real-world disorders; external validity may be limited.
Single-case experiments (single-subject design):
Focus on one participant with systematic manipulation of the IV and repeated measurements.
Baseline (A) data are collected before manipulation; then the IV is introduced (B); changes are observed; may return to baseline (A) to test reversibility (ABAB design).
Example: using rewards to reduce disruptive behavior; behavior improves when rewards are given, reverts when rewards stop, and improves again when rewards resume.
Longitudinal studies:
Track the same individuals over an extended period to observe changes and development.
Epidemiological studies:
Examine the distribution and determinants of disorders in populations; focus on prevalence and incidence across groups and time.
Practical note on quasi-designs:
These designs are often necessary due to ethical and practical constraints but generally provide weaker internal validity than randomized controlled trials.
Protecting and evaluating human participants
Human participants require careful ethical protections.
Institutional Review Board (IRB):
A committee (often five or more members) at a research facility that reviews, approves, and monitors studies involving human participants.
Possesses the power to require changes, disapprove, or stop a study if participant safety or rights are jeopardized.
In the U.S., IRBs are empowered by federal agencies such as the Office for Human Research Protections and the Food and Drug Administration.
Rights of participants and informed consent:
Participation should be voluntary.
Participants must be adequately informed about what the study entails before enrollment.
Conflicts of interest and data handling in modern research:
Reports indicate that many pharmaceutical-funded studies show favorable outcomes, while independent studies show fewer favorable results; this underscores potential bias in research sponsored by industry.
The American Psychological Association requires data sharing for replication and reanalysis, though not all authors comply (reasons include data misplacement, ethical concerns, or ongoing studies).
Ethical challenges in social media and online data:
Online research participants may differ from in-person participants (WEIRD concerns and sampling biases).
Studies have raised concerns about consent when data are public or collected through social platforms without explicit participant consent (e.g., Facebook mood manipulation study in 2014).
Debates over informed consent in digital data use and privacy concerns have led to initiatives like the Social Data Initiative (SSRC) to develop ethical guidelines.
WEIRD participants and generalizability:
More than 70% of psychology studies use college students as participants.
WEIRD: Western, Educated, Industrialized, Rich, Democratic populations; findings from WEIRD samples may not generalize to non-WEIRD populations.
WEIRD participants tend to be more educated, more individualistic, more narcissistic, and more risk-taking, among other differences, which can limit generalizability.
Internet vs in-person sampling differences:
Online (weird): ~57% female; more racial diversity; less education; older; poorer; more geographic diversity.
In-person: ~71% female; less racial diversity; more educated; younger; wealthier; less geographic diversity.
Replication and publication bias:
Replication is essential for establishing accuracy and generalizability.
A sizeable share of replication studies are unsupportive or weaker than original findings (53% unsupportive vs 47% supportive in some analyses).
Fewer replication studies are conducted over time, and negative replication findings are less likely to be published, which can distort the scientific record.
Data sharing and transparency:
While journals encourage data sharing, actual sharing rates may be low due to ethical concerns, ongoing studies, or misplacing data.
Potential ethical concerns in modern research:
Direct manipulation of social media content without informed consent can raise ethical issues and potential harm to participants (e.g., mood manipulation studies).
There is ongoing debate about privacy, consent, and the use of public data in research.
Confounds and research design integrity
Confound: a variable other than the IV that may be influencing the DV, leading to spurious conclusions.
Examples: location of the therapy office, background music, participant expectations, or other situational factors.
Strategies to guard against confounds:
Include a control group.
Use random assignment.
Implement mask (blinding) designs where possible.
Animal research and ethics:
Animal studies can provide insights but raise ethical concerns; institutions use guidelines to protect animals and require committees (e.g., IACUC) to oversee proposal and ensure humane treatment.
Rough estimates suggest between 12 \text{ to } 27\text{ million} animals used in U.S. research annually, with about 0.5\text{ million} being primates and other species; animal research has contributed to life expectancy gains and medical advances but remains controversial.
The rights and welfare of animals in research:
Public health and regulatory bodies have established guidelines to minimize harm and ensure humane treatment; the number of animals used has declined compared to previous decades.
Special case: animals in therapy and interventions:
Some studies explore animal-assisted therapy or companionship as calming interventions, illustrating the broad range of potential treatments.
Ethical and methodological synthesis
Research methods function best as a team: each approach has strengths and weaknesses.
Converging evidence from multiple methods strengthens conclusions; conflicting results indicate areas where knowledge is still limited.
Critical evaluation is essential before accepting findings: consider
Whether variables were properly controlled
Whether the sample is representative and large enough
Whether bias was minimized
Whether conclusions are justified by the data
Whether ethical standards were met
Key takeaways:
Case studies provide rich, detailed information but limited generalizability and internal validity.
Correlational studies identify associations and have strong external validity but cannot establish causation alone.
Experimental studies can establish causation and have high internal validity when well-controlled, but may face ethical and practical limits and potentially limited external validity.
Quasi-experimental designs are useful when true experiments are impractical, though they typically offer weaker causal inferences.
Protecting human participants is paramount, with IRBs and informed consent as foundational elements.
Replication, data transparency, and careful consideration of WEIRD sampling are critical for robust psychology.
Notable quotes and historical context (illustrative examples)
Misperceptions have driven scientific progress when challenged by research (e.g., Aristotle and others on gender, communications technology, cloning, etc.).
The field emphasizes that beliefs require testing in representative samples to avoid harm from incorrect theories (e.g., lobotomy as a cure for schizophrenia).
General admonition: "If we knew what it was we were doing, it would not be called research" (paraphrase of Einstein’s sentiment about scientific inquiry).
Summary comparisons and practical implications
Case studies vs correlational vs experimental:
Case studies: detailed, idea-generating, therapy-refining; limited internal/external validity.
Correlational studies: identify relationships and generalize to populations; cannot establish causation; rely on representative samples and statistical significance to assess real-world relevance.
Experimental studies: establish causation with random assignment and control of confounds; robust for understanding mechanisms but often constrained by ethics and practicality.
Ethical and methodological integration:
In practice, multiple methods are used together to build a coherent understanding of abnormal functioning.
If all methods converge on similar conclusions, confidence increases; if results diverge, further investigation is needed.
Pragmatic takeaway for exam preparation:
Be able to define each method, list its strengths and limitations, and identify typical research questions suited to that method.
Recognize the role of control groups, random assignment, and masking in experiments.
Understand the difference between statistical significance (p-value) and clinical significance (meaningful real-world impact).
Be aware of WEIRD sampling issues, online vs in-person data collection differences, and replication concerns in modern psychology.
Remember the ethical scaffolding: informed consent, IRBs, and responsible data handling, including data sharing and conflict-of-interest awareness.
Key formulas and numeric references to memorize
Correlation coefficient domain:
r \in [-1, 1]Statistical significance threshold commonly used:
p < 0.05Relationship descriptors (direction and magnitude) relate to the sign and absolute value of r (e.g., strong positive: |r| \text{ close to } 1; weak: |r| \text{ much less than } 1).
Animal usage estimate and impact (for context):
Animals used in U.S. research annually: 12 \text{ to } 27\times 10^6
Primates and other mammals: ≈ 0.5\times 10^6
Institutional review and data sharing acronyms:
IRB: Institutional Review Board
IACUC: Institutional Animal Care and Use Committee
SSRC: Social Science Research Council
Notable qualitative statements to recall:
Major disasters have served as natural experiments revealing patterns of psychological reactions (e.g., acute stress, PTSD symptoms after disasters).
The replication landscape shows a shift toward more replication studies but ongoing concerns about publishing negative or contradictory results.
Social media research raises unique ethical challenges (informed consent, data privacy, manipulation concerns).