PSYC 655 Midterm

Types of Control Groups

(Experimental Design) 1. IV checks (ex. Pilot testing; manipulation check: check data as you go) 2. Participant variables (ex. Demographics) Types: 1. No treatment control: threats to internal validity; ethical issues 2. Wait list Control Group 3. No-Contact Control Group 4. Nonspecific Treatment or "Attention Placebo" group (same level of attention, better than nothing) 5. Routine/Standard Treatment (active) control group (difficulty in determining effectiveness of treatment 6. Comparative Effectiveness Trial: use two successful treatments and compare them to see which is more successful

Threats to Internal Validity

-History -Maturation -Testing -Instrumentation -Statistical Regression -Attrition -Selection -Diffusion or Imitation of Treatments -Compensatory Equalization of Treatments -Experimenter Expectancy

Strategies that will Accelerate the Accumulation of Knowledge

-Promoting and Rewarding Paradigm Driven Research: accumulates knowledge by systematically altering a procedure to investigate a question or theory, rather than varying many features of methodology; opportunity to incorporate replication and extension into a single experimental design -Author, Reviewer, and Editor Checklists: can ensure disclosure of obvious items sometimes forgotten without being too much extra work (easy to implement) -Challenging Mindsets that Sustain the Dysfunctional Incentives: committees can invest time to examine quality, potential impact, and direction of the research agenda -Metrics to Identify what is Worth Replicating: more important to replicate findings with a high replication value because they are becoming highly influential -Crowd Sourcing Replication Efforts: feasibility could be enhanced by spreading the data collection effort across multiple labs -Journals focused on Soundness not Importance of Research -Lowering or Removing Barrier for Publication: if anyone can publish it is not longer an incentive and can change the mindset that publication is the end of the research process -ULTIMATE SOLUTION: OPENING DATA, MATERIALS, AND WORKFLOW: open data can allow the ability to confirm, critique, or extend prior research and increases ability to find and correct errors and improve reported results; open methods allow improvement of research already conducted, there are many factors that could be important but are not mentioned in an article; open workflow can clarify whether a finding was a priori or discovery in the course of conducted the research

Pros of Physiological Measures

-Psychophysiological measures generally unobtrusive and nonintrusive -data can be collected continuously -findings objective since don't depend on judgment -potential for multidimensionality of data (when data collected more than one source) -comparisons of effects of different stimuli -individual differences

Cons of Physiological Measures

-specialized and sometimes expensive equipment required -few standard protocols -expertise required -validity may be questionable due to poor signal-to-noise ration of some measures -considerable btw. and within variability -inference -sensitive to more than one factor -no generally agreed upon theoretical framework -Example: Skin Conductance (fear, measures sweat; stress)

History (Threat)

1) history - changes in DV are due to another factor occurring between pre- and post-tests • External event occurs at same time as treatment/manipulation e.g., national or local news event, initiation of new program or policy • Solution? o Control group and random assignment o Shield participants from external event o Choose DVs that would not be affected by external events

Cons of Behavioral Observations

1. Behavior needs to be accessible (can reliably observe it), behavior may be related to something in the environment (making it hard to observe outside of that), and we need to know how much data is needed, how many times, and under what conditions 2. Participant reactivity: data not representative of typical behavior 3. Observer Drift: gradual shift away from original behavior definition resulting in inaccurate recording of behavior 4. Demand Characteristics: on both sides; better to keep both participants/experimenters blind 5. Only what is actually seen and recorded is known 6. Informant discrepancies: different informants may have different conclusions; can hinder goals of therapy

Journal Review Process

1. Editor in Chief: decodes whether sufficient contributions made to existing literature 2. Associate/Action Editor: manuscript assigned depending on area expertise 3. Editorial Board: provide expertise 4. Ad Hoc Reviewer: work with associate editors

Experimenter Expectancy (Threat)

10) experimenter expectancy - believe treatment is working • Experimenter influences participants' behavior or interprets participants' behavior in direction of expected results • Solution? o Double-blind procedure o Multiple experimenters or coders

Maturation (Threat)

2) maturation - change in participant over time • Natural changes (e.g., physical, intellectual, emotional) in participants as they age • Particularly problematic for longitudinal studies and child research • Solution? o Control group

Testing (Threat)

3) testing - reactive effects of testing • Reactivity of DV to simply being in a study or assessed e.g., self-monitoring with observational measures • Repeated assessment of DV → familiarity → test bias • Practice effects • Solution? o Unobtrusive or indirect measures (less reactive) o Control group

Instrumentation (Threat)

4) instrumentation - change due to change in measurement instrument from pre to post • Use of different measure at pre-test and post-test • Solution? o Control group o Use exact same measure o Practice effects or pretest sensitization o Indirect measures

Statistical Regression (Threat)

5) statistical regression - people selected for extreme scores will have less extreme scores when tested again (solution is a control group) • Participants selected for extreme scores tend to have less extreme scores when retested • Extreme scores are rare and likely to be affected by • Response bias - satisficing • Response error • State factors (e.g., mood, fatigue) • Solution? o Use measures with high test-retest reliability o Use measures that are less likely to be affected by random error o Control group o Multiple pretests o Practice effects or pretest sensitization

Attrition (Threat)

6) attrition - when it is not random • Proportion of participant loss differs by treatment group • Particular participants more likely to withdraw from study • Solution? o Pretest experimental conditions o Incentives for study completion o Statistically control for participant factors

Selection (Threat)

7) selection - non-random assignment • Participant factors differ by group • Non-equivalent groups • Quasi-experimental design • Lack random assignment • Matching groups does not guarantee equivalence • Solution? o Use random assignment whenever possible o Assess potential covariates o Replication o Selection-history, selection-maturation, selection-instrumentation effects

Diffusion or Imitation of Treatments (Threat)

8) diffusion or imitation of treatments - treatment bleeds over into control group • Participants learn about other conditions and manipulation • Communication between participants • Within-subjects design • Sensitization - participant guesses hypothesis • Reactivity • Demand characteristics • Solution? o Ask participants not to discuss study with others until data collection complete o Limit use of within-subjects variables (e.g., mixed-factorial design)

Compensatory Equalization of Treatments (Threat)

9) compensatory equalization of treatments - control participants seek treatment when they realize their group status • Experimental groups receive unequal compensation • Differential attrition • Treatment seeking in control group • Solution? o Wait-list control group o Additional incentive for control o Active control o Placebo that causes side effects

Internal Validity

Degree to which changes in DV can be attributed to IV ◦ Extent to which possible confounds or alternative explanations are eliminated ◦ True experiment > quasi-experiment > correlational design True experiment ◦ Random assignment ◦ Manipulate IV ◦ Control extraneous variables Internal validity ◦ Generally higher priority ◦ More important for basic research ◦ Easier to attain in the lab - more control

Dos and Don'ts of Reviewing

Don't -Include your publication recommendation in the review -Rewrite manuscript for author -Redesign the study -Note all spelling or grammatical errors - copy editor's job Do -Maintain a professional and respectful tone -Offer feedback that improves the science and/or communication of the science -Educate author about literature/references missing from manuscript -Focus on big picture

Pseudoscience vs. Psychology

Pseudoscience: o No predictive power o Not based on facts or empirical evidence o Unsystematic, sloppy methods o Confirmation bias o Not logical o Contradictory, lacking in parsimony o Not peer-reviewed, results cannot be verified Scientific Method: o Ask Question o State Hypothesis o Design Research Method o Collect Data o Analyze Data o Make Inferences from the Data o Revise or Replicate o Must ask questions that can be empirically measured or tested (solvable problems) • Limited to current knowledge and research techniques o Questions must be falsifiable (Karl Popper) Able to disprove o Findings must be observed, replicated, and verified by others (public verification) • Ensures phenomena are real and observable • Makes science self-correcting • Scientific findings and explanations are TENTATIVE

Recipe for Writing Professional Reviews

Read article/manuscript/proposal/book and take notes Let simmer Write the review Let simmer again (at least overnight) Read it and edit freely Ready to Submit!

What to Include in Critical Review

Separating major from minor concerns, specific considerations for each section of manuscript Title and abstract- does it convey meaning, achieving purposes Intro- lit review that doesn't cite every study on record, good theory important Method- novice reviewers sometimes try to rewrite w/out acknowledging more than one way to complete a study, whether choice of methodology and data analytic techniques appropriate Results- presented in clear and concise manner Discussion- how and why explanations for study findings Tables and figures- should improve readability of manuscript oBe aware of review length, sensitivity when reviewing for international journals, copy editing manuscript, signing the review

Advantages of a Three-Dimensional Naturalistic Approach (Experimental Design)

a. New Empirical Laws are Discovered: not only are unexpected findings often obtained, but the findings are generally stronger, more convincing, and more highly valued by the research community; laboratories could be arranged for some natural-dimension research, but demand characteristics might have provided serious rival explanations b. Making the Research More Credible to Participants: human subjects are genuinely involved; creates high degrees of realism -Can increase internal validity: if subject is engaged in responding naturally to a real event in a natural setting, or whenever the subject is unaware or forgets he is involved in psychological research, most threats to internal validity disappear -Increasing external validity: increase generalizability; tends to get put on the back burner after internal validity; in opposition to internal validity; can be more generalizable to people, situations, etc. c. Implementing Naturalistic Methods with Experimental Controls: sacrifice experimental control to observe naturalistically; alternative to experimental control is rescheduling natural treatments and natural behaviors to occur in a lab setting; can conduct parallel studies both in lab and naturalistically

Hypothetico-deductive (HD) approach

deducing or deriving one or more explicit and testable hypotheses from some plausible theory about the phenomena of interest prior to designing one's research

Construct Validity

extent to which measure reflects accurately the variability among objects as they are arrayed on the underlying continuum to which the construct refers • Six Facets of Construct Validity: a. Content b. Substantive c. Structural d. Generalizability e. External f. Consequential • Major Problem: possibility that constructs of interest may be conceived at several different levels of meaning (construct validity at one level may not apply to validity at another level) • Second Problem: construct validity might be quite variable depending on conditions of measurement (ex. Subject wanting to flatter clinician) • Valid measure only to the extent that audience believes: g. Construct defined in satisfactory way h. Measure captures what is implied by definition i. Scores on measure related to broader phenomena implied by idea • Protections: j. Improvement in measurement k. General consensus about measurement questions l. Ways to stipulate the construct validity of specific measures for specific purposes m. Pay close attention to conditions/context in which measurement occurs n. Peer review o. Postpublication critique p. "Critical Multiplism" (don't rely too heavily on one measure) • Major threats to construct validity (fixes relevant to mock study used as example) o Inadequate explication of constructs o Construct confounding o Mono-method bias o Confounding constructs with levels of constructs o Reactive self-report changes o Reactivity to experimental situation o Experimenter expectancies • Measurement of construct validity: class and modern approaches o MTMM = establish convergent and discriminant validity, determine extend that measures share method variance o Quantifying construct validity = use effect size patterns rather than relative significance levels of correlations

Statistical Conclusion Validity

inferences about whether presumed cause and effect covary, power to detect effect in given study and appropriate use of statistics to accurately quantify effect (Extent to which cause and effect conclusions can be detected using appropriate statistical methods) • If at a given level of alpha, DV and IV covary/ correlated • Three judgments to be made about covariation = Is the study sensitive enough to detect covariation btw. variables? If it is sensitive enough, is there evidence that the IV and DV covary? If there is evidence, how strongly do the variables covary? • Major threats to statistical conclusion validity and the fixes o Low statistical power: possible solution- using a larger sample size o Violated assumption of statistical tests: possible solution- researcher needs to be aware of assumptions to ensure assumptions met in data set o Fishing and error rate problem ("fish" through data set to find significant results): possible solutions- Bonferroni correlation, false discovery rate o Reliability of measures: possible solutions- report reliability and strive for adequate reliability o Restriction of range: possible solution- avoid dichotomous outcome measures and use median-splits to divide continuous variables o Reliability of treatment implementations: possible solutions- carefully train researchers in collecting data o Random irrelevancies in experimental setting: random assignment of groups, exert control over environment o Random heterogeneity of respondents: possible solutions- if not possible to sample participants with similar characteristics, then measure characteristics and use for blocking or as covariates o Inaccurate effect size estimation: possible solution- estimates based on prior literature

Cons of Self-Report

o Bxs & attitudes strongly influenced by features of research instrument, wording, format, & context o First task-find out whether respondent's understanding matches what the researcher intended; tapping the same facet? o Two processes of question comprehension - semantic understanding (literal meaning) and pragmatic meaning (using inferences about questioner's intention) • Use simple wording with familiar, unambiguous terms • Inferences lean on conversational assumptions: • Summary - speakers should try to be informative, truthful, relevant, clear o Frequency scales and reference periods o Example: How often with "less than once a year" may lead participant to think researcher has a relatively rare event in mind o Length of reference period impacts o More intense emotions and marital disagreements in retrospective than current reports because intense experiences are more memorable • Rating scales o Have to make interpretations of what researcher means, e.g. "not at all successful" or "zero" o Negative to positive scale indicates bipolar construct o Only positive scale indicated unipolar construct • Question Context o Researcher's epistemic interest - clues provided by researcher's affiliation (often overlooked) Example - survey says "Institute for Personality Research" at top o Adjacent question - interpretation of a question's intended meaning is influenced by adjacent questions • Social desirability and self-presentations • Mundane, frequent bxs are most poorly remembered • Attitudinal measures highly context specific o In self-administered questionnaires, subsequent questions can influence previous ones because can jump around o Assimilation effects - more positive judgements when feeling positive and vice versa • Item order

HARKing can be a cost to science

o Can translate Type 1 Errors into theory: theory is constructed to account for an illusory effect; a priori hypotheses would have a theoretical foundation independent of result o There is no immediate possibility of disconfirmation o Can promote statistical abuses o Confirmation and hindsight bias • Limited exploration of existing literature o Promotes distorted model of science: presents a rosy picture that hypotheses always come true

Reliability

o Consistency or dependability of a measuring technique • Sufficient reliability ≥ .70 o Test-retest • Stable traits, characteristics, behaviors o Alternate Form o Interrater • Cohen's kappa - agreement for categorical data • Intraclass Correlation Coefficient (ICC) - more than two raters o Interitem - internal consistency • Item-total correlation (≥ .30) • Split-half reliability (Spearman-Brown formula) • Cronbach's alpha coefficient

HARKing is incentivized

o Journals want scientists to evaluate an existing testable idea, not to discover the idea through the research o Confirmation is more important to journals than disconfirmations o Journals don't want invalid approaches (waste of time) o HARKing provides a better fit to good science script and a general story script

Pros of Behavioral Observation

o Objective (if didn't know; disguised) o Systematic (coding scheme) o Continuous or time sampling o Repeated Measurement

Assessing Observer Agreement and Reliability

o Seek to gauge how close measures come to reflecting the true state of affairs o Variance due to: construct, systematic error, and random error o If two observers agree, but systematic error is high, less meaningful o Interobserver agreement has limitations but is essential o Point-by-point methods that ignore time • Percent agreement - can provide feedback to observers during training, used in older literature, does not control for chance • Cohen's Kappa - zero to one (from no to perfect agreement); results from a matrix • Weighted Kappa - allows you to regard some disagreements more seriously than others o Point-by-point methods are best for training observers, and summary measures are best for communicating with colleagues o Summary statistic measures • Simple probabilities • Conditional or transitional probabilities • Others (e.g., Yule's Q) • Intraclass correlation coefficients

How to find practice of HARKing

o Too Convenient: does not arise from another sensible theory and is confirmed by results of the study o Too good to be true: theoretical argument is reasonable, but mismatch between plausibility, coherence, and power of theory and findings o Methodological Gaff: poor fit between study design and goal (because study was not designed to test HARKed hypothesis)

Analogue Research (Experimental Design)

• Analogue research has two dimensions = may refer to a sample of characteristics or can describe research procedures, analogue across sample and procedural domains; emphasis on discovery over confirmation (ex. Using a college sample to get info then you can move on to a specific population of people) • Analogue sample = exists on continuum from clinical to nonclinical, determined by external validity/ generalizability of sample compared to clinical pop. of interest 1. Should be driven by sound theoretical analysis of primary research questions 2. Studies of individuals with preexisting diagnoses may tap characteristics and behaviors that develop over course of disorder • Analogue sample may be chosen to assess symptoms not of a clinical threshold but instead approximate number/ severity symptoms • Or use theory-based selection, considering extent which study goals concern symptom threshold • Analogue procedure = extent to which setting and instrumentation differ from context in which clinical phenomena of interest actually occurs in natural environment 1. More similar lab setting to naturalistic setting, greater ecological validity - Further considerations • Skepticism/ uncertainty about analogue research- confusion largely due to lack of clear definition of "analogue" and lack of focus on theory as guide to analogue selection and evaluation • Use of analogue research may have ethical benefits, goal of analogue research to assist in advancement and knowledge of phenomena that inspired the analogue, work to expand generalizability (systematic replication)

Sampling: Using College Students

• Can remedy by replication using other populations (which rarely happens) or estimate using known differences with the general population (which is what this article does) • Compared with older adults, college students (narrow age range - upper SES) have 1) less crystalized attitudes (social and political), 2) Less-formulated sense of self (self-esteem, identity confusion and diffusion, inadequate integration of past, present, and future selves, insecure, depressed), more egocentric 3) stronger cognitive skills, 4) stronger tendencies to comply with authority, and 5) less stable peer group relationships (strong need for peer approval, dependency, overidentification)- laboratory setting intensifies these (esp. since lab is setting of complying with authority and separating from peer group) • Resulting view of human nature based on these samples say that people... o Quite compliant, highly socially influenced o Do not rest self-perceptions on introspection o Behave inconsistently with attitudes o Emphasize cognitive processes over personality, material self-interest, emotional irrationalities, group norms and stage-specific phenomena • Current (at the time) consensus that this isn't a big deal • Problems with narrow data base o May wrongly describe strength of relationships (y=a+bx) 1) incorrectly estimate the size of b, 2) range of x's may not map onto those in ordinary life, 3) laboratory may underestimate variety of interacting conditions

Multi-Method, Multi-Informant Approach

• Different measurement types, different people (ex. Parent/child inconsistency) • Informant discrepancy • Weaknesses to each type so strengthens assessment battery (same with informant)

Experimental Design

• Differentiating Experimental and Nonexperimental Psychopathology 1. Pure experimental research: Adheres to the experimental method; a priori focus on elucidating variables contributing to origins of abnormal behavior; identification of causal variables; foundation for future research on treatment and prevention; does not INDEPENDENTLY yield complete etiology; highly localized, specific, and focused on subset of variables; use of experimental task does NOT necessarily mean that research is experimental 2. Quasi-experimental research: manipulation of independent variables and evaluation of their effects on psychological processes in samples with a well-established type of psychopathology, among persons that vary in some established psychopathology risk dimension or show subclinical features; cannot unambiguously determine cause; can be combined with pure experimental induction and then quasi-experimental conditions; can involve independent variables that exacerbate existing abnormal bx 3. Non-patient and descriptive research: No manipulation of independent variables and mostly limited to correlational questions in non-clinical samples; increasingly popular; may try to elucidate differences between clinical and non-clinical samples

Ecological Validity

• Extent to which experimental setting, measures, manipulations, etc. reflect real-world, natural situation • Mundane Realism - "match on superficial features between an experimental manipulation and some aspect of everyday life" • Experimental Realism - "impact and meaningfulness of a manipulation for the subject" o Psychological realism

Threats to External Validity

• External Validity - to what extent can findings be generalized across populations, settings, and epochs • Threats: 1) interaction of selection and experimental condition - can findings from sample be generalized to other groups? 2) interaction of setting or context and experimental condition - can findings in one setting generalize to another? 3) interaction of history and experimental condition - would current findings apply in another time period or would they apply had participants had different histories?

External Validity

• External validity has traditionally asked the question of generalizability: to what populations, settings, and so on, do we WANT the effect to be generalized? • Author argues that the question of external validity is not the same as the question of generalizability. The goal of our research might not be to generalize so the situations are perfectly in line with our targeted "real life" scenarios • Why else do an experiment other than to predict real-life behavior? o We may be asking whether something CAN happen o We may be testing something that ought to happen IN THE LAB o We may demonstrate the POWER of a phenomenon by showing that it happens even under unnatural conditions that ought to preclude it. o We may use the lab to produce conditions that have NO COUNTERPART in real life at all Extent to which findings generalize across variations ◦ situations ◦ types of participants ◦ types of manipulations ◦ different measures ◦ across time Ways to improve external validity ◦ Replication ◦ Representative sample ◦ Benchmarking - comparing lab and field studies External validity ◦ More important for applied research ◦ Easier to attain in the field ◦ Not always necessary to demonstrate ◦ depends on research question and purpose

HARKing

• HARKing (hypothesizing after the results are known): presenting post hoc hypotheses in a research report as though they were a priori hypotheses; not traditional "scientific induction"

Implicit Measures

• Implicit Association Test (IAT): assesses strength of an association between a target concept and an attribute dimension by considering the latency with which participants can employ two response keys when each has been assigned a dual meaning • Physiological approaches have also been employed as implicit measures of attitudes (facial electromyography to examine racial prejudice) • Implicit assumes that attitudes are ones for which individuals lack awareness- it is more appropriate to view the measure as being implicit not the attitude • With prejudice and stereotypes the correlations between implicit and explicit measures tend to be quite low • When motivation and/or opportunity are low, behavior is expected to be largely a function of the automatically activated attitude (the implicit measure should prove predictive) • When motivation and opportunity are high the explicit measure should be more predictive (explicit measure will have been influenced by these same motivational forces) • Example: a priming measure involving photos of fat and thin women predicted how far participants later placed their own chair from that of a fat woman • Flexible Correction Model: some individuals possess a naive theory about the biasing effects of their negativity- a theory that overestimates its influence- so they overcompensate • IAT have pursued a "known-groups" validity approach; IAT scores of two groups of individuals differ in the expected way • Future research concerning the predictive validity of the IAT may benefit from the consideration of moderation variables

Pros of Self-Report

• Inexpensive • Easy to use • Anonymity • Group administration

Sampling: Race

• It is important to provide detailed demographic information of your human participants • Argues that representative sampling is essential for external validity • If published research does not specify demographics and argues generalizability of findings, what it advocates for could be harmful to under-represented groups. • Majority of research has been conducted with predominately middle-class White Americans. • This practice reflects a cultural bias of the field (lack of journals focused specifically on people of color). • There is an implicit (and wrong) assumption that as long as people of color are adequately represented in research samples, then the results should generalize to those populations. • There are many difficulties achieving representative samples: practical and expense

Alternatives to Generalization (External Validity)

• It is worth knowing that an event CAN occur, even under restricted conditions (I.e. in a lab people judged participants wearing glasses as smarter). Saying something can affect our judgement says something about our judgement-don't need to apply specific situation to "real world". • Take a hypothetico-deductive method: this result did not happen in these circumstances which means that a theory is either false, or need qualification. • Change the way we generalize- example infants acquiring speech. Researchers can make conclusions by generalizing from sample to population, but by what happened in the sample. Show that a theory needs to be qualified appropriately. • Checklist of comparing artificial setting to real world setting is not helpful. • In each case think of: what conclusion we want to draw, and whether the specifics of our sample or setting will prevent us from drawing it.

Sampling: Types

• Randomization and random sampling can equate groups on several nuisance variables simultaneously and do not require the researcher to be aware of: a. How the important nuisance variables are related to the response measure b. The identities of the important nuisance variables c. The number of important nuisance variables • Random sampling and randomization can be expected to result in the equivalence of large samples but need not result in the equivalence of small samples • Groups Nonequivalent: if the proportion of subjects who fall in one category of the dichotomous nuisance variable in one group is at least twice that of the other group • The probability the equal sized treatment and control groups will be nonequivalent: d. Increases with an increase in the number of nuisance variables e. Generally decreases with one exception, with an increase in the size of the total subject pool • Random Sampling: instead of having a fixed number of subjects available for the study, the researcher has a very large pool of potential subjects; groups viewed as nonequivalent on one nuisance variable if the proportion of one group falling in one category of the dichotomized variable differs from the corresponding proportion in the second group by at least 0.3333 • Major consequence of nonequivalence is bias in the estimates of relative efficacy or treatment effects (may be over or underestimated) • Reversal Paradox: treatment mean will be greater than the control mean even though the control mean is greater than the treatment mean within each of the two levels of the nuisance variables • Researchers can control for nonequivalence by using large samples • If scientists believe in the law of small numbers they tend to ignore/downplay the influence of sampling error on estimates of treatment effects and therefore, tend to downplay the importance of statistical significance tests, which evaluate estimates of treatment effects in relation to estimates of sampling error • When samples are large (40 per group) randomization and random sampling appear to be effective in creating groups equivalent on maximum number of nuisance variables (moderate: 20-40 has even shown to appear to work well) o Simple Random Sampling o Simple Random Assignment o Cluster Sampling: ex. West Virginia Picture

Cons of Implicit Measures

• The few reports regarding test-retest reliability for various priming measures have ranged from abysmally low to moderate • Measurement error plays a role in the low relationship that have been observed among various implicit measures • IAT responses are considered automatic because they are expressed without intention or control, although perceivers may become aware of the attitude under scrutiny during the task • The IAT may be influenced by such environmental associations • Implicit measures also display some sensitivity to context • Depending on the nature of the priming task: lexical decision is more sensitive to stereotype activation, whereas one based on adjective connotation is more sensitive to attitude activation • Stimulus Selection (ex. Pictures chosen)

Pros of Implicit Measures

• This type of measure is likely to be free of social desirability concerns because they do not directly ask the participant for a verbal report