Psychology
Observation and Hypothesis Development
Science typically starts from an observation of the world.
This observation leads to a hypothesis: a tentative idea about whether the observation might be true for others or is unique to a case.
Testing the hypothesis is essential in science; not just speculating, but conducting experiments or measurements.
Results must be interpreted, which may show that the initial hypothesis was right, wrong in part, or completely off.
The process often loops back to new observations, refining the hypothesis and expanding inquiry.
Researchers develop several operational definitions to specify how a concept will be observed or measured in a study.
Operational definition example: measuring a complex construct like love requires a concrete, observable, and measurable specification.
A long example: allogrooming in rats is defined with specific behavioral criteria to ensure consistent coding.
Allogrooming example definition (long):
One rat grooms the other, usually around the neck and head, but may include other body parts. Grooming motions include snout contact with body part with continuous movement of the snout contact. The grooming may include rapid nibbling while the groomed rat is immobile. Movement of the groomed rat may elicit attack and kicking behaviors from the groomer. This excludes when one rat is actively pushing down upon the submissive, preventing the submissive rat from moving away.
This precision helps readers of a paper know exactly what behavior was coded as allogrooming; self-grooming and other behaviors are distinct.
Allogrooming is a common social behavior in rodents and is analogous to grooming in other species (e.g., chimps). The definition is specific to the study's measurement, but researchers can operationalize constructs in multiple valid ways.
Operationalizing a construct can be done through different methods, depending on the construct and the research questions.
Example: operationalizing anxiety
Physiological measures: heart rate changes, sweaty palms during stress tasks (e.g., mental math under time pressure).
Self-report: surveys asking about perceived anxiety in specific environments (e.g., crowds, new situations).
Behavioral performance: tasks measuring concentration under time pressure (e.g., three-minute math test with a memory component).
All three approaches aim to capture the same construct (anxiety) from different angles; this is similar to a structuralist approach—decomposing a construct into measurable parts.
No single method is inherently better; different methods can strengthen understanding if convergent results emerge across approaches.
Key takeaway: operational definitions translate abstract ideas into concrete, measurable variables, enabling replication and comparison.
Types of Research and Data Types
When designing a study, researchers decide on the type of research: descriptive, archival, correlational, experimental.
Data output can be qualitative or quantitative:
Qualitative data: non-numerical; often collected via surveys or interviews and recorded in words or non-numeric form.
Quantitative data: numerical data suitable for statistical analysis; easier to apply standard statistics.
Qualitative advantages/disadvantages:
Pros: rich context, depth, nuanced understanding.
Cons: harder to analyze with traditional statistics; time-consuming; subjective interpretation.
Quantitative advantages/disadvantages:
Pros: easier to summarize with statistics (means, variances, patterns); scalable to large samples.
Cons: may miss nuance; requires careful measurement and often predefined scales.
Operationalization and measurement choice influence data type and analyses.
Archival Studies (Secondary Data)
Archival studies use materials already gathered and publicly available; no direct interaction with participants.
Examples of archival data: census data, police reports, college transcripts, online datasets.
Advantages: can study large populations and long time spans; data already exists.
Disadvantages: limited control over how data were collected; potential biases in original data; limited ability to modify measures.
Research questions must fit the structure of the available archive.
Example scenario: test whether younger generations have more anxiety than older generations by comparing historical vs. current institutional data (e.g., mental health treatment rates).
Operationalize anxiety using archival indicators (e.g., treated anxiety cases per decade).
Limitations: changes in diagnostic criteria, reporting practices, and help-seeking behavior across eras may confound comparisons.
Descriptive Research and Reactivity
Descriptive research aims to observe and describe behavior; often thought of as “watching” but can be methodologically tricky.
Reactivity: people alter their behavior when they know they are being watched or recorded.
Example: dancers at a wedding might dance more conservatively if told they are being observed.
In animal studies, the presence of a camera can change how rodents behave (e.g., sniffing a camera before mating).
Higher reactivity lowers external validity:
External validity: the extent to which results generalize to real-life settings or other populations.
If participants alter behavior due to observation, results may not generalize well.
To mitigate reactivity in descriptive research, researchers can use naturalistic observation in real environments (unzapped by explicit observation) to improve ecological validity.
Naturalistic Observation and Participant Observation
Naturalistic observation: record naturally occurring behavior in a natural environment without intervention; aims to reduce reactivity.
Participant observation: the researcher attempts to become part of the activity, blending in as a participant.
Examples include studying cults or social groups from within (e.g., undercover study of a doomsday cult).
Advantages: behaviors are more authentic; participants may behave more naturally when unaware of precise observation.
Disadvantages: context may alter behavior; ethical concerns around consent and deception; results may not generalize across contexts or cultures.
Ethical considerations: researchers must balance scientific knowledge with the rights and welfare of participants; deception and covert observation raise debates about consent.
Case Studies
Case studies focus intensively on one person or a small number of individuals; can generate rich data and theory development.
Applied Behavior Analysis (ABA): uses case-based observations to shape and modify behavior, widely used with autism spectrum disorders.
Phineas Gage: classic neuroscience case showing that frontal lobe damage led to dramatic personality changes (impulse control, decision-making) despite survival.
Early publications by case studies catalyzed discoveries about brain-behavior relationships.
Other neuroscience case studies: rare brain injuries revealing regional brain functions (e.g., language and executive functions).
Limitations of case studies:
External validity is low because findings may not generalize beyond the single case.
Replication is challenging; establishing causality from a single case is difficult.
Verifying effects requires multiple cases or triangulation with other methods.
Examples of famous cases and extensions (e.g., Phineas Gage) illustrate how single cases can inform broader theory.
Surveys and Psychological Tests
Surveys are a common descriptive technique: fast, scalable, and cost-effective; often use questionnaires or Likert-type scales.
Population vs sample considerations:
Target population: the entire group the researcher wants to draw conclusions about.
Sample: subset of the population actually studied.
Random sampling improves external validity by increasing representativeness.
Convenience samples (e.g., a single class) limit generalizability.
Social desirability bias: respondents may answer in a way they think is socially acceptable rather than truthfully.
Researchers use lie scales or embedded checks to detect inconsistent or biased responding.
Surveys pitfalls demonstrated by pizza preference poll:
Question design can influence results (e.g., not specifying pizza type, size, gluten-free options, or sample differences).
Small samples and unrepresentative populations limit external validity.
Population scope and inclusion criteria (e.g., gluten-free needs) affect generalizability.
Psychological tests: standardized assessments (e.g., depression scales, achievement tests, aptitude tests, intelligence tests, personality tests).
Used for diagnosis, treatment planning, and evaluating treatment efficacy.
Not all tests are equally valid for every purpose; context and norms matter.
Diagnostic processes often rely on checklists and surveys to quantify symptom severity and impairment.
Statistics: Describing and Inference
Statistics are used to assess and evaluate patterns in data and to determine if observed patterns are due to chance or reflect true effects.
Central tendency measures:
Mean: ar{x} = rac{1}{n}
abla
a sum{i=1}^n xi
Note: correct LaTeX for mean is ar{x} = rac{1}{n} ext{since the transcript uses a simple average}
Median: middle value when data are ordered.
Mode: most frequent value.
Variability measures:
Range:
Standard deviation: s =
( rac{1}{n-1}
ar{(x_i - ar{x})^2}
)Visual intuition: a bell-shaped distribution centers around the mean; high variability yields a flatter shape; low variability yields a tighter cluster around the mean.
Inferential statistics: used to decide whether sample results generalize to a larger population and whether observed differences are unlikely due to randomness.
Correlational Research
Correlation measures whether two variables vary together and to what extent they are related.
The correlation coefficient r ranges from -1 to +1:
r = +1: perfect positive relationship
r = -1: perfect negative relationship
r = 0: no linear relationship
Example relationships:
Years of education (x) and salary (y): usually a positive correlation (as education increases, salary tends to increase).
Class absences (x) and exam scores (y): typically a negative correlation (more absences, lower scores).
Intelligence (x) and shoe size (y): often near zero correlation (no meaningful relationship).
Important caveat: correlation does not imply causation. A third variable or confounding factors may drive the relationship.
Third-variable examples: hardworking or resource-rich individuals may influence both education and salary; missing class could be associated with其他 factors like motivation or study habits.
Fun real-world correlations (examples provided for illustrative purposes): cheese consumption and accidental strangulation by cheese; spelling bee word length and spider-related deaths; etc. These illustrate spurious correlations and the danger of inferring causality from correlation alone.
Experimental Research
Experimental research actively manipulates an independent variable to observe its effects on a dependent variable, enabling causal inferences.
Key terminology:
Independent Variable (IV): the variable the experimenter deliberately changes.
Dependent Variable (DV): the outcome measured and expected to change as a result of the IV.
Confounding variables: uncontrolled variables that could influence the DV, threatening internal validity.
Participant variables: individual characteristics of participants (e.g., mood, prior experience) that may influence responses.
Situational variables: environmental factors during the study (e.g., time of day, noise) that may influence responses.
Distinguishing from descriptive/archival/correlational work: experimental research directly tests causality by controlling and manipulating conditions.
Example: sleep and test grades
Hypothesis: students who sleep a full night perform better on exams than sleep-deprived students.
Variables to identify: IV, DV, potential confounds; control procedures to isolate sleep as the causal factor.
Design considerations for experiments:
Random assignment to conditions helps equate groups on participant variables.
Control groups and experimental groups to compare outcomes.
Managing confounds to ensure observed effects are due to the IV.
Ethical considerations in manipulating sleep, stress, or other sensitive factors.
Important Concepts for Exam Prep
Operational definitions are critical for replicability and clarity.
Reactivity reduces external validity in descriptive research; naturalistic observation can mitigate this.
Archival data provide large-scale insights but limit control over data collection methods.
Descriptive research describes behavior but does not infer causation.
Correlation identifies associations but not causation; beware third variables.
Experimental research establishes causation through deliberate manipulation and control of variables.
Distinguish between qualitative and quantitative data; choose methods accordingly.
Understand limitations of case studies regarding external validity and replication.
Remember key statistics concepts: mean, median, mode; range; standard deviation; and inferential statistics for generalization.
Population vs sample: random sampling improves generalizability; convenience samples limit external validity.
Social desirability bias can distort survey data; strategies include lie scales and validity checks.
When evaluating studies, consider: construct validity (are we measuring what we intend?), internal validity (are there confounds?), external validity (do results generalize?), and reliability (are measurements consistent?).
Quick Formulas and Concepts to Memorize
Mean: ar{x} = rac{1}{n}
abla ext{sum}{i=1}^n xiMedian: middle value in ordered data
Mode: most frequent value
Range:
Standard deviation: s =
( rac{1}{n-1}
ar{sum}{i=1}^n (xi - ar{x})^2)Correlation coefficient: r = rac{
ar{x}
ar{y}}{
ar{…}}More complete form of the correlation (standard form):
r = rac{ extstyle rac{ ext{cov}(X,Y)}{(n-1)}}{ extstyle sX sY} = rac{ ext{Cov}(X,Y)}{sX sY} = rac{
extstyle rac{1}{n-1}
ar{ igl(X - ar{X}igr)igl(Y - ar{Y}igr)}}{sX sY}Note: In practice use the standard definition: r = rac{ extstyle rac{
ar{(X - ar{X})(Y - ar{Y})}}{
}}{ ext{(SD)}X ext{(SD)}Y}Exploration of external validity: generalizability across different populations and settings
Replication is essential for verification of effects in case studies and experiments