Notes on The Scientific Method, Ethics, Descriptive Methods, Correlation vs Causation, Experimental Methods, Data Interpretation, and Critical Evaluation (Sections 2.1–2.7)

2.1 The Scientific Method

  • Ebbinghaus’s quote: psychology has a long past but a short history. The field asks timeless questions about the mind, but psychology as a science is relatively young.
  • The long past vs short history contrast:
    • Long past: questions about mind and behavior date back to philosophers (e.g., Aristotle).
    • Short history: psychology did not become a rigorous science until relatively recently; science uses empirical methods, not just rational argument.
  • Rationalism vs empiricism:
    • Rationalism: knowledge comes from reason and logical argument rather than experience.
    • Aristotle’s view (4th century BCE): thoughts, perceptions, and emotions were products of the heart; the brain and other organs were secondary and served to cool the blood to maintain rational state (Figure 2.2 reference).
    • Modern understanding: logic alone cannot explain human behavior; empirical methods (experimentation) are required.
  • Why science is needed in psychology:
    • Our intuitions and common sense often fail to predict behavior in real situations (e.g., people often don’t notice identity changes during conversations; Simons & Levin, 1998).
    • In a change-blindness study, 53% of participants failed to notice a switch in conversation partner in one study and 67% in another, illustrating limits of perception and common sense.
    • These findings motivate the use of controlled, empirical methods to form and test theories.
  • Psychological science as a discipline:
    • We gain new knowledge by using experimental methods to test theories and build models of behavior.
  • The Kitty Genovese case as a springboard:
    • 1964 murder in NYC where many witnesses did not help; initially thought to reflect a breakdown of social bonds in big cities.
    • Darley and Latané argued against this and proposed the diffusion of responsibility and the bystander effect as explanations for why people don’t help when others are present.
    • This case led to formal investigation of bystander behavior and the scientific method’s application to social psychology.

The Scientific Method in Psychology

  • Identify the problem: curiosity about why many witnesses did not help in Kitty Genovese case; basis can come from observation, prior research, theory, or intuition.
  • Gather information: review literature and theories (e.g., diffusion of responsibility, bystander effect).
  • Develop a hypothesis: more bystanders → less individual helping (bystander effect).
  • Design and conduct an experiment: create a study to test the hypothesis (e.g., diffusion of responsibility manipulated via number of bystanders in a simulated emergency).
  • Restart the process: consider replication, replication with extension, or new lines of inquiry.
  • Analyze data and draw conclusions: determine whether data support predictions; acknowledge that findings are not absolute proof; remain open to new evidence.

2.2 Research Ethics

  • Historical motivation: the Tuskegee syphilis study (1932–1972) highlighted ethical violations and spurred formal ethical guidelines.
  • Tuskegee Syphilis Study: >600 African-American men recruited with promises of meals, medical treatment, and burial insurance; researchers followed disease progression without treating participants even after treatment became available.
  • Institutional Review Boards (IRBs): federally required for research at institutions receiving federal funding; composed of diverse faculty and staff; evaluate studies for ethics and safety.
  • What IRBs look for:
    • Sound research design.
    • Risks minimized and reasonable in relation to potential benefits.
    • Benefits outweigh risks.
    • Informed consent and voluntary participation; the ability to withdraw at any time without consequence.
    • Safeguards to protect well-being.
    • Data privacy and confidentiality.
  • Informed consent: participants are informed about the study’s purpose, procedures, risks, benefits, data protection, and rights; participation is voluntary; participants can ask questions; consent can be withdrawn.
  • Informed consent questions (example from Quiz): Informed consent means that participants give permission to participate after all details and risks are known; participants retain the right to withdraw.
  • Deception in research:
    • Sometimes deception is used to preserve authentic behavior, but four criteria must be met:
      1) The research poses no more than minimal risk to participants.
      2) There is justification that deception is necessary and no feasible alternative exists.
      3) The benefits of deception outweigh potential harms and trust is preserved.
      4) A debriefing occurs after the study, and participants are informed about deception and can ask questions; consent may be re-obtained for data usage.
  • Debriefing and consent re-obtainment after deception:
    • Participants are told about the deception, the study’s true purpose, and can withdraw their data if they wish.
  • Real-world ethics example: the Facebook Emotional Contagion study (Kramer, Guillory, & Hancock, 2014) explored emotional contagion via manipulated news feeds in 689,000+ users without explicit informed consent; raised significant ethical debate about consent in online research.
  • Standards for deception approval (4 criteria) are designed to protect participants and maintain public trust in psychology research.

2.3 Descriptive Methods

  • Descriptive research aims to describe what is, not why it is.
  • Four common descriptive methods:
    • Naturalistic observation
    • Participant observation
    • Case studies
    • Surveys
  • Naturalistic Observation
    • Observation of behavior in natural environments without manipulation.
    • Qualitative (descriptions, notes) or quantitative (counts) data.
    • Strengths: generates ideas and captures behavior in real-world contexts.
    • Limitations: no control over variables; replication difficult; observer bias; potential lack of interrater reliability (consistency between observers).
    • Interrater reliability: agreement among observers, crucial for reliable observations.
  • Participant Observation
    • Researcher joins the group under study to gain access and deeper understanding.
    • Potential benefits: rich, contextual data and access to daily life.
    • Limitations: observer influence on group dynamics; bias; researchers may misinterpret behaviors due to immersion.
  • Rosenhan (1973) – a classic case study using participant observation in clinical settings
    • Eight healthy researchers (pseudopatients) admitted to 12 psychiatric hospitals using false identities and reports of hearing voices (words like “empty,” “hollow,” “thud”).
    • Once admitted, they behaved normally; clinicians diagnosed most as schizophrenic; patients sometimes suspected the pseudopatients were faking.
    • Outcome: clinicians showed a bias toward false positives; healthy individuals could be misdiagnosed as mentally ill, challenging the validity of psychiatric diagnoses at the time.
    • Conclusion: clinical assessment can be biased toward labeling healthy individuals as ill; highlights the need for improved checks and balances in diagnosis.
  • Rosenhan video discussion (Review item): In the video, the pseudopatients were discharged with a diagnosis of schizophrenia in remission (this reflects the enduring bias and diagnostic practices of the era).
  • Case Studies (HM – Henry Molaison)
    • HM underwent surgical removal of hippocampi to treat intractable seizures; post-surgery, he could not form new memories, though other cognitive functions remained.
    • This case provided foundational evidence for hippocampus involvement in memory formation and helped identify distinct memory systems (episodic, semantic, procedural).
    • Strengths of case studies: detailed, in-depth data about rare phenomena.
    • Limitations: limited generalizability; time and expense; may not represent broader populations.
  • Surveys
    • Efficient way to gather data on opinions, attitudes, or behaviors from a sample.
    • Population vs. sample: population is the whole group of interest; sample is a subset selected for study.
    • Sampling error: differences between the sample and population; biased samples can lead to inaccurate conclusions.
    • Question wording can bias results (e.g., “Should your college allow speeches that might incite violence?” vs. “Should your college forbid speeches on campus that might incite violence?”; the only change is one word, but responses may differ due to framing).
    • Response bias and social desirability: participants may respond in socially acceptable ways or misreport behavior to look better.
    • Kinsey studies (1948, 1953): large-scale surveys of sexual behavior that revealed differences between social attitudes and actual practices, helping spark cultural shifts during the sexual revolution of the 1960s.

2.4 Correlation, Not Causation

  • Correlation measures the relationship between two variables; how one changes with the other.
  • Visual representation: scatterplots show two variables on axes; a positive correlation moves upward from left to right; a negative correlation moves downward from left to right; zero correlation shows no clear pattern.
  • Correlation coefficient r:
    • Direction and strength of the relationship are conveyed by r, which ranges from −1extto+1-1 ext{ to } +1.
    • The sign indicates direction (positive or negative); the magnitude indicates strength.
    • The strength is determined by the absolute value ∣r∣|r|; a larger absolute value indicates a stronger relationship.
    • A perfect positive or negative correlation has ∣r∣=1|r| = 1; zero correlation has ∣r∣o0|r| o 0.
  • Important caveat: correlation does not imply causation.
    • Observed correlations can be due to a third variable (a confounding variable).
    • Example: number of dogs in a city and number of fire hydrants might be positively correlated due to population size (a confound) rather than one causing the other.
    • Population size is a plausible confound that can explain many spurious correlations.
    • Informatively, more correlations exist as examples (e.g., margarine consumption and divorce rates) but do not imply causation; many correlations arise from complex, unseen variables.
  • Best approach to establish causation: controlled experiments with random assignment and manipulation of the independent variable (IV).

2.5 Experimental Methods

  • Goal: identify cause-and-effect relationships.
  • Key elements of experiments:
    • Hypothesis: a clear, testable prediction about how variables relate;
    • Independent Variable (IV): the manipulated factor (e.g., type of video game – violent vs non-violent);
    • Dependent Variable (DV): the measured outcome (e.g., aggressive behavior);
    • Experimental group: receives the treatment (violent video game);
    • Control group: does not receive the treatment (non-violent game or neutral task);
    • Random assignment: participants are randomly assigned to groups to equalize confounding variables (e.g., gender, age, background);
    • Confounding variables: other factors that could influence the DV; must be controlled or balanced between groups to ensure internal validity.
  • Operational definitions: define precisely how variables will be measured (e.g., aggression could be operationalized by the amount of hot sauce added to a drink after exposure to a video game).
  • Hypothesis example (violent video games):
    • A testable, falsifiable hypothesis might be: "People who play violent video games will exhibit more aggressive behavior than people who do not play violent video games."
    • An operationalization of aggression could be the amount of chili sauce the participant permits to be given to a taste tester.
  • Random assignment and control groups enhance internal validity, enabling stronger causal inferences about the effect of the IV on the DV.
  • The role of confounds: equal distribution of gender and other potential confounds across groups helps ensure observed effects are due to the IV.
  • Internal validity: the extent to which observed effects are due to the manipulation of the IV and not other factors.
  • External validity: the extent to which findings generalize to real-world settings.
  • Four key ideas in experimental design:
    • At least two groups (experimental and control);
    • Clear IV and DV;
    • Random assignment to conditions;
    • Measurement of the DV after manipulation; comparison of group outcomes.
  • Deception in experiments (again): may be used if necessary to preserve the study’s authenticity; must meet the four criteria outlined in 2.2 and be followed by thorough debriefing.
  • The opening example: how to structure a simple experiment on the effect of violent video games on aggression and how to map confounds and ensure randomization.

2.6 Making Sense of the Data

  • Describing data: central tendency and variability
  • Central tendency:
    • Mean: the average (sum of scores divided by the number of scores).
    • Median: the middle value when data are ordered.
    • Mode: the most frequently occurring value.
    • The mean is sensitive to outliers; the median can provide a better central tendency when outliers are present.
  • Example of outliers: a data set with nine incomes around 42{,}000$-$60{,}000$ and one outlier income much higher would pull the mean upward, while the median remains a more representative central value. ( ext{outlier}
    ightarrow ext{extreme value})
  • Describing data: spread (variability)
    • Range: difference between the highest and lowest values.
    • Standard deviation (SD): average distance of data points from the mean; a measure of dispersion.
  • Inferential statistics (a brief intro):
    • Inferential statistics help determine whether observed patterns are likely to reflect true effects in the population, not just in the sample.
    • Statistical significance is a probabilistic threshold used to decide whether to reject the null hypothesis.
  • Statistical significance:
    • If the probability of an observed result under the null hypothesis is very small, we deem the result statistically significant.
    • Common threshold: p < 0.05((p<0.05 is often used to claim a result is unlikely due to chance).
  • Probability intuition:
    • Example with cards: in a standard deck, half the cards are red: rac{26}{52} = 0.5;drawing10cardsyieldsanexpectednumberofredcardsaround; drawing 10 cards yields an expected number of red cards around5; getting 9 or 10 red cards in a row would be highly unlikely (low probability). A rough calculation shows that very long red-card streaks are rare, which informs judgments about what is “uncommon” or statistically significant.
  • Practical concerns in statistics:
    • Not all data lead to definitive conclusions; researchers must consider sampling, measurement quality, and potential biases.
  • The testing effect (Roediger & Karpicke, 2006):
    • Hypothesis: repeated testing improves long-term recall more than repeated studying.
    • Experimental design: participants read a passage; one group is tested on recall (retrieval practice), the other group re-reads for extra study.
    • Results: immediate recall may be slightly better for the study group, but after delays of days to a week, the testing group recalled substantially more information (about 15 ext{%} more) in long-term recall.
    • Conclusion: recall-based testing enhances long-term retention and demonstrates a causal effect of testing on memory.

2.7 Critical Consumer of Information (Spotting Fake News and Misleading Research)

  • The challenge of evaluating information in a world with abundant data and online content.

  • Strategies for evaluating credibility and reliability:

    • Be wary of correlational studies presented as causation; identify whether the study design supports causal conclusions.
    • Check for objectivity: what does the author stand to gain? Is there potential bias or vested interest?
    • Distinguish fact from opinion; assess whether arguments rely on anecdotes or systematic data.
    • Examine how variables are operationalized and whether measurements are valid and reliable.
    • Look for whether conclusions are appropriate for the study design (correlational vs. experimental).
  • Additional notes on correlation vs causation and interpretation:

    • Simple correlations may be interesting but do not prove that one variable causes another.
    • Consider potential third variables or confounds (e.g., population size in the dog-fire hydrant example).
  • Quick reference glossary:

    • Correlation: associated with, related to, corresponds with.
    • Cause-and-effect: causes, results in, creates, influences, affects.
  • Figure 2.64: Strategies for spotting fake news (visual checklist provided in the chapter).

  • Final cautions:

    • Be deliberate and systematic when evaluating research claims.
    • Practice and familiarity improve the ability to spot weak designs, biases, and inappropriate causal inferences.
  • Key formulas and numbers echoed throughout the notes:

    • Correlation coefficient range: -1 \, \le \, r \, \le \, 1
    • Absolute strength of correlation: |r|(strengthincreasesas(strength increases as|r| \to 1)
    • Significance threshold: p < 0.05
    • Card probability example: rac{26}{52} = 0.5
    • Memory recall testing results (Roediger & Karpicke, 2006): long-term recall advantage for testing condition ~15\%$$ over the control group after delays.
  • Important historical anchors mentioned in the text:

    • Aristotle and rationalism vs empiricism; Ebbinghaus’s characterization of psychology as having a long past but short history.
    • Kitty Genovese case as a catalyst for diffusion of responsibility and bystander effect research.
    • Tuskegee syphilis study and the development of IRBs and ethical guidelines.
    • Rosenhan (1973) and HM (Henry Molaison) cases as foundational descriptive and case-study examples in psychology.
    • Kinsey reports as early large-scale surveys influencing social attitudes toward sexuality.