EPPP common mistakes
EPPP · Understand the traps and how to avoid them.
On power and error questions, selecting 'decrease alpha' or 'use a two-tailed test' as ways to increase power, and mixing up which error (Type I vs Type II) matters more in a described applied scenario.
Candidates memorize 'power = 1 minus beta' as a formula but never map the direction each design lever moves alpha, beta, and power, so a plausible-sounding option wins.
What to do instead
Draw the 2x2 decision matrix (reality x decision) and write a one-line arrow chart: alpha down = power down; N up = power up; effect size up = power up; one-tailed = power up; measurement error down = power up. Recite the arrow chart at the start of every stats study block and answer one applied 'which error is worse here' scenario per session (e.g., screening for a treatable disease vs. executing a prison sentence).
On screening and diagnostic accuracy items, choosing the most sensitive test to 'rule in' a condition, or computing/choosing a predictive value without using the stated base rate.
The mnemonic 'sensitivity rules out' is memorized as a phrase without the underlying logic, and candidates have never actually worked a predictive value problem with a low prevalence, so they default to the intuitive but wrong answer.
What to do instead
Memorize SnNout (high Sensitivity, negative result, rules out) and SpPin (high Specificity, positive result, rules in). Then work one full example from scratch weekly: sensitivity .90, specificity .90, prevalence 1 percent, 10,000 people gives about 990 true positives and 990 false positives, so PPV near 50 percent; repeat with prevalence 20 percent to see PPV rise. Always locate the base rate in the stem before answering.
On ethics vignettes, selecting the most dramatic option (immediately terminating, breaching confidentiality to a family member, reporting a past crime) instead of the least intrusive adequate step, and failing to verify whether a law actually mandates the action.
Harms described in vignettes are emotionally salient, and dramatic options feel like 'doing something,' while consultation, documentation, and clarification steps feel passive. Candidates also conflate general ethical duties with jurisdiction-specific mandatory reporting statutes.
What to do instead
Apply a fixed three-question ladder to every ethics item before reading options: (1) Does a specific standard or law address this exact situation, and what does it require? (2) What is the least intrusive action that still protects the client and third parties? (3) What consultation and documentation are required first? Eliminate any option that skips a rung. Drill standards 1.02 (conflicts between ethics and law), 4.05 (disclosure), and 10.10 (termination) until you can state the required sequence from memory.
Attributing SSRI therapeutic effects to the immediate rise in synaptic serotonin, or describing benzodiazepines as 'increasing the amount of GABA' rather than as GABA-A receptor modulators.
Psychopharmacology is studied by drug class and clinical indication (antidepressant, anxiolytic) rather than by molecular target, so timing and mechanism details never get encoded.
What to do instead
Build one table with columns: class, molecular target and action, time course to therapeutic effect, key side effects and withdrawal risk. Fill in at least SSRIs, SNRIs, bupropion, mirtazapine, benzodiazepines, and antipsychotics. Be able to explain why SSRIs take two to four weeks despite immediate transporter blockade (autoreceptor desensitization and downstream neuroplastic changes).
Mislabelling scenarios with neighboring social-cognitive constructs, for example calling a self-versus-other difference the fundamental attribution error, or calling a frequency-based judgment the representativeness heuristic.
Constructs were learned as isolated labels and definitions with no contrast cases, so any vignette mentioning 'blaming the person' triggers the most familiar term.
What to do instead
For each construct write a one-line discriminator keyed to the judgment: fundamental attribution error (target is another person, cause is dispositional), actor-observer (target differs: self vs other), self-serving bias (target is self, outcome valence differs), availability (ease of retrieval), representativeness (similarity to prototype), base-rate neglect (ignoring prevalence). Add your own counterexample for each pair and test yourself by covering the labels.
Interpreting a child's behavior using adult norms, such as treating a four-year-old's egocentric speech as a thought disorder, or calling a seven-month-old's distress at strangers an attachment disorder.
Clinical training and vignette practice skew adult, and milestone ages are remembered vaguely, so normal developmental phenomena get pathologized.
What to do instead
Memorize a dated developmental ladder and check the age before choosing any diagnosis: social smile 6 to 8 weeks, stranger anxiety 6 to 9 months, walking 9 to 15 months, parallel play in toddlerhood, theory of mind around age 4. Add a column for 'what looks like a symptom but is normal at this age' and review it before each lifespan practice set.
Selecting a treatment by the clinician's preferred orientation or by broad diagnosis label rather than by the empirically supported treatment matched to the disorder, age, and setting (e.g., psychodynamic therapy for pediatric OCD, or an adult-only protocol applied to a preschooler).
Familiarity and theoretical loyalty drive recognition, and the EPPP stem often describes a textbook presentation that superficially fits several orientations.
What to do instead
Memorize a compact empirically supported treatment map: OCD and specific phobia to exposure with response prevention, panic disorder to CBT with interoceptive exposure, PTSD to prolonged exposure or cognitive processing therapy, borderline personality disorder to DBT, bulimia to CBT-E, enuresis to bell-and-pad. Add a column noting where child evidence diverges, such as required parent involvement. Answer each practice item by naming the disorder, the age group, then the protocol, in that order.
Interpreting a client's behavior strictly through majority-culture norms, for example diagnosing a recent immigrant's wariness or role-restricted behavior as social anxiety or dependent traits, or recommending assimilation, without assessing acculturation, language of interview, and cultural meaning of the symptoms.
Training is Western-normative, vignettes reward textbook symptom counts, and culture is treated as background color rather than as data that changes base rates and thresholds.
What to do instead
Before choosing a diagnosis in any cross-cultural vignette, run the DSM-5 Cultural Formulation Interview headings: cultural identity, cultural explanations of the illness, cultural stressors and supports, and the clinician-client relationship. Explicitly check acculturation level, preferred language, whether an interpreter was used, and whether the behavior is normative in the client's reference group. Eliminate any option that treats difference as deficit.
Confusing standard score conversions and interpreting them, such as calling a T-score of 40 'average' or treating a 95 percent confidence interval as the band containing 95 percent of scores.
The normal curve was memorized as a picture with no practice converting between z, T, scaled scores, and percentiles, and the confidence interval is confused with the population distribution.
What to do instead
Memorize one anchor row (z = 0, T = 50, scaled score = 100, percentile = 50; plus one SD gives T = 60, scaled = 115, percentile = 84) and re-derive all other values from it. State the confidence interval definition in your own words as the range of plausible true scores around an observed score, not a range of observed scores. Do two conversion drills daily.
Misidentifying threats to internal and external validity in design vignettes, for example choosing history when the change is maturation, or claiming random assignment controls for experimenter bias.
Threat labels are memorized as a list without a design exemplar attached, so the closest-sounding term gets selected.
What to do instead
Build a threat table keyed to design type: maturation (developmental change over time), history (external event), regression to the mean (selection on extreme scores), testing (practice), instrumentation (rater or measure drift), attrition, selection. For each, write a one-sentence mini-study with IV, DV, and control, and note which design (pre-post single group, randomized posttest only, quasi-experimental) is vulnerable. Then practice classifying stems by naming the design first, the threat second.