Chapter 2 Notes: Research Methods - Thinking Critically With Psychological Science

The Need for Psychological Science

  • Psychology uses research to separate uninformed opinions from examined conclusions, helping us answer questions like how to be happier, healthier, and more successful, and how to improve relationships.
  • Science vs. speculation: psychological science relies on observation and analysis to test ideas.

The Roadblocks to Critical Thinking

  • Three roadblocks highlighted:
    • Hindsight bias: after learning an outcome, the tendency to believe one would have foreseen it (the I-knew-it-all-along phenomenon).
    • Overconfidence: humans tend to think they know more than they do.
    • Perceiving patterns in random events: people see order in randomness (apophenia).
  • Examples and implications:
    • Hindsight bias: statements like
    • "They were not a good match" after a breakup;
    • crediting coaches for a win or blaming them for a loss;
    • perceiving obvious outcomes after events such as wars or elections.
    • Overconfidence: Tetlock study with over 27,000 expert predictions showed average confidence ~80%, but accuracy <40%.
    • Perceiving order in random events: people see streaks or faces in random sequences (moon faces, grilled cheese, etc.).

The Scientific Method and Description

  • The scientific method is a self-correcting process of evaluating ideas through observation and analysis.
  • Core principle: a theory is supported if data align with predictions; if predictions fail, revise or reject the theory.

Constructing Theories

  • A theory is an integrated set of principles that explains observations and predicts behaviors/events (beyond mere speculation).
  • Examples: Evolution, sleep and memory relationships.
  • A theory should summarize a body of observations and provide a framework for understanding new data.
  • A theory should lead to testable predictions (hypotheses).

Hypotheses and Testable Predictions

  • Hypothesis: testable predictions derived from a theory.
  • Good hypotheses specify what results would support the theory or disconfirm it.
  • Example: Hypothesis — "When deprived of sleep, people will remember less from the day before."
  • Research steps:
    • Design experiments to compare memory performance after a normal night vs. a shortened night.
    • Assess whether results support or challenge the theory.

Operational Definitions and Replication

  • Operational definitions: precise, carefully worded statements of the procedures used to measure or manipulate variables.
    • Example: Sleep deprivation may be defined as "X hours less than the person’s natural sleep."
  • Why they matter: allow replication by other researchers with different participants, materials, and settings.

Criteria for a Useful Theory

  • Organizes observations.
  • Implies predictions usable to test the theory or derive applications.
  • Stimulates further research, potentially leading to revised theories.
  • Theories can be refined through multiple research approaches:
    • Descriptive methods (case studies, surveys, naturalistic observations)
    • Correlational methods (assessing relationships between variables)
    • Experimental methods (manipulating variables to observe effects)

Descriptive Methods

  • Case Study: in-depth examination of a single individual or group to reveal universal principles.
    • Examples: brain damage studies; insights into children’s minds; animal intelligence.
    • Intensive case studies can guide further study but may mislead if atypical (e.g., smokers dying younger vs. subgroups).
  • Naturalistic Observation: observation of behavior in natural settings without intervention.
    • Examples: observing chimpanzee groups, parent-child interactions across cultures, patterns of seating in schools, social media mood analysis.
    • Pros: reveals behavior in real-life contexts.
    • Cons: may miss confounding factors; cannot establish causation.
  • Notable naturalistic findings:
    • Humans laugh ~30 times more in social contexts than alone.
    • Happiest people engage in meaningful conversations rather than small talk; they may prefer talking to tweeting.
    • Pace of life studies show faster pace in Japan and Western Europe; slower in less-developed economies.
  • Limitations: naturalistic observation cannot explain why behaviors occur because many variables co-vary.

The Survey

  • Descriptive technique to assess self-reported attitudes/behaviors of a group.
  • Key feature: uses a representative, random sample of the group.
  • Examples:
    • Americans reporting more happiness than worry on a given day.
    • Percentages believing in alien beings among 22 countries.
    • Global belief in the importance of religion.
  • Wording effects: how questions are framed influences responses.
    • Examples:
    • Support: "Aid to the needy" vs. "welfare".
    • Preference for censorship vs. freedom of information depending on phrasing.
    • Framing of gun laws as "gun safety" vs. "gun control".
  • Random Sampling: frequently not possible to sample entire group; aim for a representative sample.
    • Random sample: every member has equal chance of participating.
    • Example: selecting every nth student from a list.
    • Critical thinking: evaluate who the sample represents before accepting survey findings.

Correlation and Experimentation

  • Correlation: measures how two factors vary together and how well one predicts the other.
    • Examples: twin intelligence, exam scores predicting career achievement, stress and disease.
    • Correlation coefficient: a statistical index from −1.00-1.00 to +1.00+1.00 describing the strength and direction of a relationship.
  • Key terms:
    • Variable: anything that can vary and be measured.
    • Scatterplot: graph of two variables; slope indicates relationship direction; scatter magnitude indicates strength.
    • Positive correlation: both variables rise together.
    • Negative correlation: one rises while the other falls.
    • No correlation: no predictable relation.
  • Correlation does not imply causation:
    • Example: mental illness and smoking are correlated, but this does not indicate which causes which or whether a third factor is involved.
    • Possible explanations include: smoking causes mental illness, mental illness increases smoking, or a third variable (e.g., stress) influences both.
  • Regression toward the mean and illusory correlation:
    • Illusory correlation: perceiving a relationship where none exists.
    • Regression toward the mean: extreme results tend to move toward the average on retesting; helps explain why sensational findings or superstitions arise.

The Experimental Method

  • Experimental manipulation isolates cause and effect by
    • Manipulating one or more factors (independent variables).
    • Observing the effect on a behavior or mental process (dependent variable).
    • Random assignment of participants to experimental and control groups to control for preexisting differences.
  • Independent variable (IV): the factor being manipulated.
  • Dependent variable (DV): the outcome measured.
  • Confounding variables: other factors that may influence results; random assignment helps control them.
  • Placebo effect and shaping bias:
    • Placebo effect occurs when participants improve due to expectations about treatment, not the treatment itself.
    • Placebo-controlled designs and double-blind procedures help isolate true treatment effects.
  • Double-blind procedure: both participants and researchers are unaware of group assignments to minimize bias.
  • Examples:
    • Breastfeeding study: 17,000 Belarus newborns assigned to breastfeeding promotion vs. standard care; IQ later higher by about six points in the breastfeeding group.
    • Placebo studies show improvements in athletes and mood when perceived treatment is given, even if inert.
  • Experimental validity and ethics:
    • Internal validity: confidence that observed effects are due to the manipulated variable.
    • External validity: generalizability to other contexts or populations.
    • Ethical considerations limit what can be manipulated; some variables cannot be ethically studied via manipulation.

Research Design and Ethics in Psychology

  • Research Design choices depend on the question, time, budget, and ethical constraints.
  • Common designs:
    • Descriptive: case studies, naturalistic observations, surveys.
    • Correlational: examine relationships between variables without manipulation.
    • Experimental: manipulation with random assignment to establish causality.
    • Twin studies, longitudinal studies, cross-sectional studies.
  • Ethical safeguards for humans and animals:
    • Informed consent: participants should be informed about the study and voluntarily consent.
    • Minimize harm and discomfort; protect confidentiality.
    • Debriefing after participation, including disclosure of deception if used.
    • Animal research: guidelines emphasize humane care, minimize suffering, housing standards, and regulatory oversight.
  • Debates and responsibilities:
    • Milgram’s obedience experiments highlighted ethical concerns about stress and deception.
    • Values influence what researchers study, how they study it, and how results are interpreted.
    • Language choices reflect and shape our attitudes (e.g., calling someone a conflict vs. a personality trait).

Values in Psychology

  • Values influence topics of study (e.g., productivity vs. morale, sex discrimination vs. gender differences).
  • Observation and interpretation can be biased by researchers’ values.
  • The language used to describe people and behaviors can reflect societal values and affect interpretation.

Statistical Reasoning in Everyday Life

  • The need for statistics: critical thinking requires applying simple statistical concepts to avoid misreading data and spreading misinformation.
  • Be wary of big, round numbers and undocumented estimates (e.g., 10%, 1 million).
  • Descriptive statistics summarize data; inferential statistics generalize from samples to populations.

Descriptive Statistics

  • Measures of central tendency:

    • Mode: most frequently occurring value.
    • Mean: arithmetic average, ext{Mean} = rac{
      ext{sum of all scores}}{n} = rac{

    }{n} (conceptual form).

    • Median: middle value when data are ordered.

  • Measures of variation:

    • Range: extRange=extmax−extminext{Range} = ext{max} - ext{min}

    • Standard deviation: how much scores deviate from the mean.

    • Population sd: ext{SD} = \sigma =
      rac{1}{N}
      igg(

      igg)^{1/2}

    • Sample sd: s = igg( rac{1}{n-1}
      extstyleigg(

      igg)
      igg)^{1/2}

  • Skewness: distribution symmetry; skewed distributions (e.g., income) have tails longer on one side.

  • Normal distribution: bell-shaped curve; most scores cluster around the mean.

    • About 68% fall within one standard deviation of the mean: P(|X-

    |

    ) \,\approx\, 0.68

  • Bimodal distributions have two peaks.

Inferential Statistics

  • Inferential statistics allow generalizing from sample data to a population and estimating the probability of observed differences being true rather than due to chance.
  • Key concept: statistical significance and probability of results occurring by chance.
  • Common threshold: significance often assessed at p < 0.05\, (5\%)
  • Large samples can produce statistically significant results even if effects are trivially small (example: Facebook study with large sample showing a tiny behavioral effect).

When Is an Observed Difference Reliable?

  • Reliability factors:
    • Representativeness of the sample (population represented).
    • Less variability within groups leads to more reliable differences.
    • More cases (sample size) improve reliability.
  • Significance vs. importance:
    • A result can be statistically significant but practically trivial; significance does not imply large or important effects.

Key Formulas and Concepts (Recap)

  • Correlation coefficient: r ∈ [−1,+1]r \,\in\ [-1, +1]
    • Perfect positive: r=+1.00r = +1.00
    • Perfect negative: r=−1.00r = -1.00
    • No relationship: r=0.00r = 0.00
  • Mean: xˉ=1n ∑<em>i=1nx</em>i\bar{x} = \frac{1}{n} \,\sum<em>{i=1}^n x</em>i
  • Standard deviation (sample): s=1n−1∑<em>i=1n(x</em>i−xˉ)2s = \sqrt{\frac{1}{n-1} \sum<em>{i=1}^n (x</em>i - \bar{x})^2}
  • Probability threshold for significance: p < 0.05
  • Descriptions of data types:
    • Descriptive statistics summarize data (central tendency, variability).
    • Inferential statistics generalize beyond the sample to the population.

Real-World Relevance and Ethics

  • Psychological science informs everyday decisions and public policy, from education to health campaigns.
  • Ethical considerations are central to research design, including animal welfare and human participant protections.
  • Researchers’ values can influence topics, framing, and interpretation; transparency and replication are essential for scientific integrity.