Chapter 14: Research Methods and Statistics Reference

MCAT passages live or die on study design. Every passage tells you what kind of study was run - and many questions turn on whether you can name the design. This section walks through every design you might see.

Cross-Sectional Study

Definition. Data collected from a group at a single point in time.

Example. A researcher surveys 2,000 people today and asks about their current diet and current blood pressure.

Cross-sectional studies are good for prevalence (“how common is X right now?”) and cheap to run. They are bad for causation: you cannot establish which variable came first.

Cohort Study

Definition. A group of people sharing a common characteristic (the “cohort”) is followed over time to see who develops an outcome.

Two variants based on direction in time:

• Prospective cohort. Identify the cohort today, then follow it forward. E.g., enroll 5,000 healthy adults and measure their diet now, then track who has a heart attack over the next 20 years.
• Retrospective cohort. Identify the cohort using records that already exist, then look at outcomes that have already happened. E.g., pull 30 years of employment records for a chemical plant, then check cancer registries.

Contrast with cross-sectional. Cohort studies follow people over time. Cross-sectional is a snapshot.

Longitudinal Study

Definition. Any study that repeatedly measures the same subjects over time.

Example. A child development study that tests the same 500 children every year from ages 5 to 25.

Contrast with cohort. All cohort studies are longitudinal. But not all longitudinal studies follow a specifically defined cohort - the term is broader. On the MCAT, “longitudinal” emphasizes repeated measurement over time; “cohort” emphasizes a group followed together.

Case-Control Study

Definition. An observational design that starts with outcome, then looks backward for exposure. Two groups are identified: people who have the disease (cases) and people who don’t (controls, otherwise similar). Their histories are compared to identify exposures that differ.

Example. Researchers find 100 people with pancreatic cancer and 100 matched controls without it. They ask both groups about decades of coffee consumption, to see if coffee drinkers are overrepresented among the cases.

Contrast with prospective cohort. Prospective cohort: start with exposure, wait for outcome. Case-control: start with outcome, look back at exposure. Case-control is faster and cheaper for rare diseases but more vulnerable to recall bias.

Clinical Trial / Randomized Controlled Trial (RCT)

Definition. An interventional study where the researcher assigns participants to a treatment or a control group. In a randomized controlled trial, assignment is by chance.

Example. 1,000 patients with depression are randomly assigned to either a new drug or a placebo. After 12 weeks, symptom scores are compared.

RCTs are the gold standard for causal inference because randomization balances known and unknown confounders between groups. Still, they are not always feasible or ethical (you cannot randomly assign smoking).

Experimental Study

Definition. A study in which the researcher manipulates an independent variable to see its effect on a dependent variable, with random assignment.

An RCT is one kind of experimental study. Lab-based experiments (e.g., a psychologist manipulating the gender of a name on a résumé) are another.

Quasi-Experimental Design

Definition. An experimental-style study that lacks random assignment. The researcher still manipulates an independent variable (or a naturally occurring one), but participants are not randomly assigned to conditions.

Example. A school district adopts a new math curriculum in three schools and compares outcomes to three neighboring schools that kept the old curriculum. Since the researcher didn’t randomize, the schools may differ in ways that confound the result.

Contrast with true experiment. True experiments randomize. Quasi-experiments don’t, which weakens causal claims.

Factorial Design

Definition. An experimental design that manipulates two or more independent variables simultaneously, with every combination represented. Notation like “3 × 2” means three levels of the first variable crossed with two levels of the second, producing six treatment conditions.

Example. Testing a drug at low, medium, and high doses (3 levels) with either morning or evening administration (2 levels). Total = 6 conditions.

Factorial designs allow researchers to check for interactions - does the effect of one variable depend on the level of another?

Observational Study

Definition. A study where the researcher observes without manipulating any variable. Participants are not assigned to conditions.

Cross-sectional, cohort, and case-control studies are all observational. Experiments are not.

Embedded Field Study

Definition. A qualitative research approach where observers pose as participants or embed themselves in a real-world setting.

Example. A sociologist works undercover as a waitress for six months to study the tipping norms of restaurant customers.

Retrospective vs. Prospective Chart Review

• Retrospective chart study. Examines medical records that already exist, looking backward at events already recorded.
• Prospective chart study. Sets up in advance what will be recorded going forward, and examines the charts as they accumulate.

The MCAT does not require you to calculate statistics. It requires you to read a result and know what it does and doesn’t say. This section covers every statistical concept that has appeared on the exam.

Correlation

Correlation measures the strength and direction of a linear relationship between two continuous variables. It makes no assumption about which variable causes which.

• Correlation coefficient (r) ranges from -1 to +1.
  • r = +1. Perfect positive correlation: as one variable rises, the other rises proportionally.
  • r = -1. Perfect negative correlation: as one rises, the other falls proportionally.
  • r = 0. No linear relationship.
• Strength. The closer |r| is to 1, the stronger. |r| = 0.9 is strong; |r| = 0.2 is weak.
• A coefficient outside [-1, +1] is impossible. If you see r = -1.4, it is a computational error.

Example. Height and weight in adults correlate around r = 0.7 - a strong positive correlation, but not perfect.

Regression

Regression models a directional relationship: one variable is assumed to depend on (or be predicted by) another.

• Linear regression. Fits a straight line through data, expressing the dependent variable as a function of the independent variable.
• Unlike correlation, regression assumes a one-way influence of the predictor on the outcome.

Example. Predicting blood pressure from sodium intake. Sodium is the independent variable; blood pressure depends on it.

Contrast with correlation. Correlation is symmetric (r between X and Y is the same as r between Y and X). Regression is directional.

Chi-Square Test

Chi-square (χ²) test compares two or more distributions of categorical data. Asks whether observed frequencies in each category differ from what would be expected by chance.

Example. Do men and women smoke at different rates? Count smokers and non-smokers in each group. Chi-square tests whether the two 2 × 2 distributions differ.

Use chi-square when both variables are categorical.

t-Test

t-test compares mean values of a continuous dependent variable between two groups or a group and a specific value.

• One-sample t-test. Compare a group’s mean to a known value.
• Two-sample (independent) t-test. Compare means of two different groups.
• Paired t-test. Compare two measurements on the same subjects (before/after).
• One-tailed t-test. Tests for a difference in a specific direction.
• Two-tailed t-test. Tests for a difference in either direction.

Example. Do patients on drug A have lower cholesterol than patients on drug B? Two-sample t-test.

ANOVA

ANOVA (analysis of variance) extends the t-test to three or more groups. Like a t-test, it compares means of a continuous variable across categorical groups.

Example. Compare mean test scores across four teaching methods.

If ANOVA finds a significant difference, post-hoc tests determine which groups differ.

The Null and Alternative Hypotheses

• Null hypothesis (H₀). The default assumption: there is no effect or no relationship between variables.
• Alternative hypothesis (H₁ or Hₐ). The proposal: there is an effect.

Statistical tests try to reject the null. You never “prove” the alternative directly; you only accumulate enough evidence against H₀ to make it implausible.

Type I and Type II Errors

• Alpha (α). Probability of a Type I error; conventionally set at 0.05. You are willing to accept a 5% chance of claiming an effect when none exists.
• Beta (β). Probability of a Type II error. Typically you want β ≤ 0.20.
• Power = 1 - β. The probability of detecting a real effect when one exists. Larger samples increase power.

Confidence Intervals and Confidence Level

Confidence interval (CI). A range that is likely to contain the true population parameter.

Confidence level. The percentage of such intervals, computed across repeated samples, that would contain the true parameter. A 95% confidence level implies that 95% of similarly constructed intervals would cover the true value.

Example. A study reports a mean weight loss of 5 kg with a 95% CI of [3, 7]. If we repeated the study many times, 95% of resulting intervals would contain the true population mean.

Variance

Variance. The average squared deviation of values from their mean. A measure of how spread out data is. Its square root is the standard deviation.

Low variance → values cluster near the mean. High variance → values are spread out.

The 68-95-99.7 Rule

For a normal distribution:

• 68% of values fall within 1 standard deviation of the mean.
• 95% fall within 2 standard deviations.
• 99.7% fall within 3 standard deviations.

Example. IQ has a mean of 100 and SD of 15. Then 68% of people score 85–115, 95% score 70–130, and 99.7% score 55–145.

Regression to the Mean

Regression to the mean. When an initial measurement is extreme, the next measurement on the same subjects tends to be closer to the population average - simply because extreme values often include random noise that won’t repeat.

Example. Athletes featured on the cover of Sports Illustrated often have a poor next season. Not because the magazine is cursed, but because making the cover requires an extreme recent performance, and subsequent performances regress toward the athlete’s true mean.

This is a frequent MCAT trap in passages about treatment effects: if the sickest patients were chosen for treatment and they later felt “better,” some of that improvement may just be regression to the mean, not the treatment.

Percent Change

• Doubling = 100% increase.
• Halving = 50% decrease.
• Increase from 20 to 30 = 50% increase (the base is 20, the increase is 10, $\frac{10}{20}$ = 50%).
• Decrease from 30 to 20 = 33% decrease (the base is 30, the decrease is 10, $\frac{10}{30}$ ≈ 33%).

Percent change is asymmetric - going up 50% and then down 50% does not return you to the start. (100 → 150 → 75.)

Validity asks: does this test actually measure what it claims to measure? The opposite, a test that is unreliable or measures the wrong thing, is at best useless and at worst misleading. MCAT passages regularly name a type of validity and expect you to know what it means.

Validity vs. Reliability

Before the types: do not confuse validity with reliability.

• Validity. Accuracy. Is the test measuring the right construct?
• Reliability. Consistency. Does the test give the same result on repeat measurement?

A broken bathroom scale that reads 10 pounds too high every time is reliable but not valid. A scale that gives you a different number each time, centered on your true weight, is valid on average but not reliable. You need both.

Internal Validity

Internal validity. The extent to which a study supports a causal conclusion within its own sample. High internal validity means that the observed effect is due to the manipulated variable, not to a confound.

Example. A drug trial is highly internally valid if randomization balances patient characteristics, dosing is consistent, and blinding prevents expectancy effects.

Threats to internal validity: confounding variables, selection bias, history effects, maturation, attrition.

External Validity

External validity. The extent to which the study’s findings generalize beyond the specific sample, setting, and conditions tested.

Example. A depression drug tested only on college-aged white men in one clinic may have strong internal validity but weak external validity because the sample is narrow.

Two subtypes are specifically named:

Population Validity

Population validity. Can the sample’s results extrapolate to the larger population? Requires a representative sample.

Example. A political poll with 200 respondents, all recruited from one Twitter account, has poor population validity for predicting nationwide voting.

Ecological Validity

Ecological validity. Do the study’s conditions resemble real-world settings? A laboratory memory task where people memorize word lists for money may measure something very different from how memory operates in everyday life.

Example. A driving safety study that uses a simulator has weaker ecological validity than one that uses actual cars in real traffic.

Construct Validity

Construct validity. The extent to which the test actually captures the abstract concept (the “construct”) it claims to measure.

Example. A test of “leadership” made entirely of vocabulary questions has poor construct validity - vocabulary is not what leadership is.

Two sub-types appear on the MCAT:

Convergent Validity

Convergent validity. The test correlates well with other established measures of the same construct.

Example. A new intelligence test should correlate strongly with the WAIS. If it doesn’t, it probably isn’t measuring intelligence.

Discriminant (Divergent) Validity

Discriminant validity. The test does not correlate with measures of unrelated constructs.

Example. A new depression inventory should NOT correlate strongly with a measure of extraversion. If it does, it may be confounding mood with personality.

Convergent and discriminant validity are paired: good construct validity means converging with similar measures AND diverging from unrelated ones.

Content Validity

Content validity. The extent to which the test covers the full range of the construct it claims to measure.

Example. A driver’s license exam that only tests parking has poor content validity, because driving also includes highway speeds, merging, and navigation. A content-valid exam samples across the whole domain.

Content validity is evaluated largely by expert judgment about domain coverage, not by statistical correlation.

Face Validity

Face validity. At a glance, does the test look like it measures what it claims to?

Example. A math test full of arithmetic problems has face validity for math ability. A math test that asks about your favorite color does not.

Face validity is the weakest form of validity - it is based on surface appearance only. A test can have high face validity and poor construct validity (e.g., a “stress questionnaire” that just asks about recent bad weather). But face validity matters for participant buy-in: if the test doesn’t look relevant, people may not take it seriously.

Criterion Validity

Criterion validity. The extent to which a test correlates with an outcome (“criterion”) it should predict.

Two sub-types, distinguished by when the criterion is measured:

Concurrent Validity

Concurrent validity. The test and the criterion are measured at the same time, and they correlate well.

Example. A new rapid depression screener is given to patients at the same time as the long-form BDI-II. If the two scores correlate strongly, the screener has concurrent validity.

Predictive Validity

Predictive validity. The test, given now, correlates with a criterion measured later.

Example. SAT scores taken in 12th grade correlate with college GPA measured 4 years later. The SAT has predictive validity for college performance.

Contrast concurrent and predictive. Concurrent = same-time benchmark. Predictive = future outcome. Both are types of criterion validity.

Reliability is consistency. A test is reliable when repeated measurements give similar results - across administrations, across raters, across alternate forms, or across items. A reliable test can still be wrong (reliably measuring the wrong thing is possible), but an unreliable test is unusable regardless.

Test-Retest Reliability

Test-retest reliability. The same test is administered to the same subjects on two occasions, and the scores are correlated. High positive correlation = high test-retest reliability.

Example. A personality inventory given to 200 adults in January and again in March. If the January and March scores correlate at r = 0.90, the test is highly reliable.

Traits should be stable; states should not. High test-retest reliability is expected for stable traits (IQ, personality). For mood measures, you might expect lower test-retest correlation because mood itself changes.

Inter-Rater Reliability

Inter-rater reliability. Two or more raters measure the same phenomenon; their ratings are correlated.

Example. Two judges score the same 50 gymnastics routines. If their scores agree closely, inter-rater reliability is high.

This matters whenever human judgment is involved - diagnosis, interviews, behavioral coding, grading of essays. Low inter-rater reliability means the measurement depends on who is measuring, which is a serious problem.

Intra-Rater Reliability

Intra-rater reliability. The same rater scores the same materials on two occasions. Measures whether a rater is consistent with themselves over time.

Parallel-Forms Reliability (Split-Half Method)

Parallel-forms reliability. Two versions of a test, designed to measure the same construct, are given to the same people. If scores correlate strongly, the forms are reliable.

Split-half method. Divide the items of a single test in half (e.g., odd vs. even items) and correlate the two halves. High correlation = high internal consistency between halves.

Example. The SAT has multiple “forms” administered across test dates. Parallel-forms reliability tells you that a student’s score wouldn’t dramatically depend on which form they happened to take.

Internal Consistency

Internal consistency. All items on a test should measure the same construct. If they do, scores on individual items should correlate with each other and with the total.

Cronbach’s alpha (α). The standard measure of internal consistency. Ranges from 0 to 1.

• α > 0.9 = excellent
• α > 0.7 = acceptable for research
• α < 0.6 = questionable

A test with a single dimension should have high α. If α is low, the test may be measuring multiple unrelated things at once.

Reliability vs. Validity, Once More

Reliability is necessary but not sufficient for validity. A test cannot be valid without being reliable, but it can be reliable without being valid.

Bias is any systematic error that pushes results away from the truth. Bias is distinct from random error (which averages out) - bias has a direction. MCAT passages routinely name a bias and expect you to know which it is and how it would distort the result.

Selection Bias vs. Sampling Bias

Selection bias. Participants are not assigned to conditions in a way that produces comparable groups. Often an issue in quasi-experimental designs.

Example. Patients who choose to try a new treatment may differ systematically from those who don’t (more motivated, sicker, richer). Comparing their outcomes mixes treatment effects with selection effects.

Sampling bias. The sample is drawn from the population in a way that overrepresents some members and underrepresents others.

Example. A political phone poll conducted only during weekday business hours will overrepresent retirees and people who don’t work a traditional job.

Contrast. Selection bias is about group assignment; sampling bias is about who ends up in the sample in the first place. They often appear together.

Attrition Bias

Attrition bias. Participants drop out of a long study non-randomly - those who stay are systematically different from those who leave.

Example. A 10-year weight-loss study where participants who regain their weight quit, while those keeping it off remain. The final sample looks healthier than the starting cohort, distorting results.

Social Desirability Bias

Social desirability bias. People answer questions in ways they believe will be viewed favorably, not truthfully.

Example. Surveying college students about how much they cheat, drink, or use recreational drugs. They underreport.

Mitigation: anonymity, indirect questions, behavioral measures instead of self-report.

Self-Report Bias / Subjective Bias

Self-report. Any method where a participant reports their own attitudes, feelings, or behaviors - surveys, questionnaires, interviews.

Self-report bias (subjective bias). All self-report is vulnerable to the fact that people may misremember, misinterpret the question, exaggerate symptoms, or underreport to minimize problems.

Example. Patients may exaggerate pain to justify opioids, or underreport pain to seem tough. Either way, the self-report measurement is systematically off.

Self-report is still extremely useful - it is cheap, fast, and sometimes the only way to access internal experience. But it should not be mistaken for objective measurement.

Reconstructive Bias (Memory)

Reconstructive bias. Memories are not recordings; they are reconstructed each time they are retrieved, and the reconstruction is influenced by current beliefs, schemas, and suggestions.

Example. Eyewitnesses asked leading questions (“Did you see the broken headlight?”) often report a broken headlight even when none existed. Research by Loftus has repeatedly shown that memory can be altered by the wording of post-event questions.

Hindsight Bias

Hindsight bias (the “knew-it-all-along effect”). After an event happens, people systematically overestimate how predictable it was in advance.

Example. After a stock market crash, commentators claim the signs were obvious. Before the crash, almost no one was selling.

Normalcy Bias (“Can’t Happen to Me”)

Normalcy bias. The tendency to underestimate the probability and severity of disasters, because “things have always been normal.”

Example. Residents who refuse to evacuate before a hurricane because they’ve never been hit before.

Implicit Bias

Implicit bias. Unconscious attitudes or stereotypes that affect understanding, judgment, and behavior, without awareness.

Example. A physician who is consciously egalitarian but nonetheless spends less time with patients of a different racial background, without noticing the pattern.

Measured by tools like the Implicit Association Test (IAT), which times how quickly subjects pair concepts - faster pairing suggests a stronger unconscious association.

Cognitive Bias (Umbrella Term)

Cognitive bias. Any systematic deviation from rational judgment. Examples include confirmation bias, availability heuristic, anchoring, representativeness heuristic. Covered more fully in Chapter 4 (Memory, Attention, and Cognition).

Demand Characteristics and the Good-Subject Tendency

Demand characteristics. Cues in an experiment that tell participants what the experimenter is looking for. Participants often then comply, distorting the result.

Good-subject tendency. Participants act in line with what they believe the experimenter wants, specifically to be helpful.

Participant’s role demands. Participant expectations of what the experiment requires of them.

Example. A subject in a mood study might laugh more readily than normal because they intuit that the researcher is studying positive mood.

Mitigation: blinding, deception, cover stories, unobtrusive measures.

Response Bias Family

Several biases in how people respond to self-report items:

• Leading questions. Question wording that pushes toward a specific answer. “How much did you enjoy our excellent service?” biases upward.
• Response rate. People who respond to a survey often differ systematically from those who don’t. Low response rates threaten sampling validity.
• Ambiguity of questions. Vague items produce inconsistent answers and unreliable data.
• Acquiescence bias. Tendency to agree with statements regardless of content.
• Extreme response bias. Tendency to use the endpoints of a scale rather than the middle.

Reactivity and the Hawthorne Effect

Reactivity. People change their behavior when they know they are being observed.

Hawthorne effect. A specific form of reactivity: workers in a factory lighting study improved their performance regardless of whether lighting was raised or lowered - the improvement came from being watched, not from the intervention.

Mitigation: naturalistic observation, hidden observation (with ethical clearance), long-term observation that allows participants to habituate.

Observer / Experimenter Bias

Experimenter bias. The researcher’s expectations influence observations, measurements, or participant behavior. Mitigated by blinding - the experimenter doesn’t know which participants are in which condition.

Rosenthal effect / Pygmalion effect. Classic demonstration: teachers told certain (randomly selected) students were “gifted” unintentionally treated them differently, and those students really did perform better. Expectations shape reality.

This section covers the specific research techniques, control conditions, and procedural concepts that appear in MCAT passages without necessarily fitting into any of the earlier buckets.

Qualitative vs. Quantitative

• Quantitative. Data as numbers. Amenable to statistical analysis. “73% of respondents reported…”
• Qualitative. Data as descriptions, narratives, themes. Interviews, ethnographies, content of open-ended answers. Yields depth at the cost of statistical generalizability.

Most rigorous research programs use both (“mixed methods”).

Data Collection Methods

Survey

Survey. A structured set of questions asked to a sample of respondents. Can be in person, by phone, online, or by mail.

• Closed questions. Fixed answer choices (multiple choice, Likert scale). Quantitative.
• Open questions. Free text; participant answers in their own words. Qualitative.
• Semi-structured interview. Pre-set questions plus follow-ups based on the responses - a middle ground.

Content Analysis

Content analysis. A method for summarizing content (text, images, speeches, social media posts) by systematically coding and counting features.

Example. Coding 500 newspaper articles about immigration for the presence of specific frames (“economic threat,” “humanitarian concern,” “cultural enrichment”) and comparing coverage across publications.

Network Analysis

Network analysis (social network analysis, SNA). Investigates social structures using networks and graph theory. Nodes are people (or organizations); edges are relationships.

Example. Mapping Twitter follower relationships to identify influential users or communities.

Participant Observation

Participant observation. The researcher participates in the activities they are studying, observing from the inside.

Example. An anthropologist lives in a village for a year, joining daily life while recording observations. Extends beyond passive naturalistic observation because the observer is a “player” in the action.

Participant observation introduces reactivity - the observer’s presence may influence the behavior being observed - but gives access to context no outside observer could obtain.

Operationalization

Operationalization. The process of translating a fuzzy, abstract concept into a specific, measurable variable.

Example. “Depression” is an abstract concept. An operational definition might be “a score of 20 or higher on the Beck Depression Inventory-II.” Now the concept is something a study can actually count and compare.

Good operationalization is concrete, reliable, and faithful to the underlying concept (good construct validity). Bad operationalization creates constructs that have nothing to do with the abstract idea they claim to measure.

Counterbalancing

Counterbalancing. A procedural technique to control for order effects - changes in the dependent variable due to the sequence of conditions rather than the conditions themselves.

Example. In a within-subjects study, half the participants receive condition A then B; half receive B then A. This way any practice, fatigue, or carryover effect is balanced across both orders.

Order effects include practice effects (improvement with repeated trials) and fatigue effects (worsening with repeated trials).

Types of Control Conditions

• Positive control. A treatment expected to produce a response - used to confirm that the experimental system can detect the effect when present.
• Negative control. A condition expected to produce no response - used to confirm that the observed response is specifically due to the manipulation and not to extraneous factors.
• Vehicle (vehicular) control. The delivery medium minus the active ingredient. In a drug study, the vehicle control is the solution (saline, oil, etc.) given without the drug - anything the vehicle alone might do is accounted for.

Example. A drug study might use: a positive control (a known effective drug), a negative control (no treatment), and a vehicle control (saline injection). Only by comparing the experimental drug against all three can the researcher isolate its effect.

Direction of Relationships

• Unidirectional relationship. Cause → effect. One variable influences the other, not the reverse.
• Reciprocal relationship. Cause ↔ effect. Each variable influences the other. Common in social processes: self-esteem and academic success, marital quality and mental health.
• Reciprocal exchange. A social interaction characterized by back-and-forth giving.

Reciprocity and Free Riders

Reciprocity. A social rule: we repay what another person has given us. Gifts, favors, and invitations generate an expectation of return.

Example. If a colleague covers your shift, you feel obligated to return the favor. This obligation sustains ongoing exchange relationships.

Free riders. Individuals who benefit from a group’s resources without contributing. The problem: they erode the reciprocity that holds cooperation together. Most groups punish free riders - sometimes at costs exceeding the harm done - because failing to punish invites more free riding. Individuals often work hard to avoid being seen as free riders.

Tests of Association and Psychophysics

A handful of specific tests appear by name in MCAT passages:

• Word association test. Participant produces the first word that comes to mind in response to a stimulus word. Used to probe semantic networks, emotional associations, and implicit memory.
• Implicit association test (IAT). Measures the strength of automatic associations between concepts by timing how quickly participants sort paired stimuli. Faster sorting = stronger implicit association. Widely used to detect implicit biases that self-report would miss.
• Operational span testing (OSPAN). Dual-task working memory measure: participants remember a list of items (words, digits) while performing a simultaneous processing task (math problems). Predicts fluid intelligence better than simple digit span because it taxes the central executive. (See Chapter 4.2.)

Psychophysics

Psychophysics. The branch of experimental psychology that quantitatively investigates the relationship between physical stimuli and the sensations and perceptions they produce.

Example. Weber’s law (the JND is a constant proportion of the base stimulus) comes from psychophysical studies of how large a change must be before it is noticeable. (See Chapter 1.)

Standardized Tests

Standardized test. A test where all takers answer the same (or equivalent) questions in the same way, scored in a consistent manner. Enables comparison of relative performance across people.

Example. The MCAT itself. SAT, GRE, and the Stanford-Binet IQ test are other standardized tests.

Standardization is key for large-scale assessment but has been criticized for cultural bias and narrow measurement of capability.

Ethics of Studies

Research on humans is governed by several ethical principles:

• Informed consent. Participants must understand what the study involves and voluntarily agree.
• Confidentiality and anonymity. Data is handled to protect identity.
• Right to withdraw. Participants can leave at any time without penalty.
• Minimization of harm. Physical, psychological, and social risks must be minimized and justified by scientific value.
• Debriefing. After the study (especially if deception was used), participants are fully informed about the study’s purpose.
• IRB review. An Institutional Review Board evaluates proposals before the study begins.

Deception is sometimes permitted for scientific reasons, but requires strong justification and thorough debriefing. Classic pre-IRB studies (Milgram obedience, Zimbardo’s Stanford Prison Experiment) could not be approved today in their original form.