L2 Flashcards
5 methodological issues in experiments
reliability, file-drawer problem, validity, WEIRD subjects, experimenter bias
Reliability
consistent results across multiple occasions, with different experimenters, across different labs/field sites
Inter-rater reliability vs test-retest reliability
- different raters, same results
- same kid, different time, same results
File-drawer problem
many studies don’t get published and sit in (virtual) file drawers as positive results are more likely to get published
Validity
having a measure that accurately reflects the process or construct of interest
e.g. looking time and preference
Internal validity
whether the observed effects can be attributed to what you’re measuring (dependent variable) and your manipulation (independent variable) without a confound
External/ecological validity
whether or not the findings are generalizable to the general population (real world) or just the sample (lab)
Potential solution to the WEIRD bias
ManyBabies projects where dozens or hundreds of labs around the world do the same study, estimate the size of the effect, and probe cross-cultural differences
Attrition issues in infant studies
babies leaving the sample
2 necessary conditions in conducting infancy research
measurement equivalence and control for everything to prevent confounds (e.g. parental influence)
measurement equivalence: the same construct is being measured across groups or across time
Examples of recruitment methods in infancy research
birth records, hospital recruitment, buying mailing lists, facebook ads, posting flyers
5 kinds of research designs
correlation, experiment, quasi-experiment, naturalistic observation, structured observation
Correlation
values on 2 or more variables are observed and the relationship between them is assessed
Benefits and limitations of correlational design
- can evaluate the direction and strength of the relationship between variables that may even be impossible or unethical to manipulate
- cannot draw causal conclusions (direction-of-causation problem) and third-variable problem
Experiment
- participants are randomly assigned to conditions or groups
- experimenter manipulates the independent variable/s and measures the dependent variable/s
Benefit vs limitation of experimental design
can make causal conclusions but some variables cannot be manipulated
Between-subjects vs within-subjects experiments
- comparison between experimental group and control group
- each subject gets both experimental and control items
Which kind of experimental design has more statistical power?
within-subjects
because individual differences are eliminated
Quasi-experimental designs
natural variables (e.g. sex, age, SES) are treated as groups so members are not randomly assigned
Limitations of quasi-experimental designs
- essentially correlational so difficult to conclude causality as third variables may be in play
- need to ensure groups match as closely as possible on other variables
2 main designs for studying development (change over time)
Cross-sectional vs longitudinal design
- groups of different children of different ages are compared
- the same children are examined over time at different ages (collected at intervals of months or years)
Benefits and limitations of cross-sectional design
- relatively fast and requires less commitment from families
- statistically less powerful due to potential individual differences and cohort differences
cohort differences: age differences are confounded with group differences
Microgenetic design
type of longitudinal design wherein the same children are observed repeatedly after short intervals (e.g. daily, weekly) over a relatively short time period as they master a task (e.g. onset of crawling, first word learning)
Benefits vs limitations of longitudinal design
- better information on developmental change and individual differences are not as influential
- expensive, time-consuming, cohort effects are masked, test-retest issues (e.g. practice effects)
Cohort-sequential design
- combination of cross-sectional and longitudinal designs
- different cohorts of children are tested over time with each starting the study at different ages
Age-held constant design
all children studied are the same age but vary in some way that influences or contributes to development
e.g. exposed to one language vs multiple languages
Methods to measure infant preference
preferential looking, conditioned head-turn task, preferential reaching, high-amplitude sucking
Habituation
inducing a negative preference (or boredom) for a stimulus through repeated exposure, resulting in dishabituation to a novel stimulus
Comparator model
explains habituation
we preferentially orient to stimuli that doesn’t match our internal representations
Violation of expectation
exposure to stimuli that is more/less interesting at baseline (without inducing a negative preference) will elicit a response from infants
Imitation
a link between an infant’s own actions and those of others (e.g. neonatal imitation)
How can an infant’s memory be tested?
deferred imitation paradigms
Examples of physiological/neural measures
- autonomic nervous system activity (e.g. heart rate, blood pressure, stress hormones)
- pupil dilation (measure of surprise/interest/concern) measured via eyetracker
heart rate slows when interested and increases when afraid
Pros vs cons of physiological/neural measures
- automatic and relatively objective
- requires special equipment and training, has a debatable reason for response, and can be overhyped
Hemodynamic response
a neurological measure
- a measure of oxygenated blood flow that gives you a sense of which brain areas are active/passive with different stimuli/tasks
- measured using MRI and fNIRS (functional near-infra-red spectroscopy)
- Pros: non-invasive, good spatial resolution, can see big structural changes over time, and some functional activity (with fNIRS)
- Cons: only a proxy for brain activity and has poor temporal resolution
Hemodynamic response
a measure of oxygenated blood flow that gives you a sense of which brain areas are active or passive with different stimuli or tasks
MRI vs fNIRS
when used to measure hemodynamic response
- only performed when asleep, has a loud noise, and no movement allowed
- sleeping not required, no loud noise, movement allowed
EEG and Event-related potential (ERP)
a measure of electrical activity in the brain, at rest or following an event
Pros vs cons of EEG and ERP
- good temporal resolution, non-invasive, and measures actual brain activity
- poor spatial resolution