research methods

pilot studies

A pilot study is a small scale version of the investigation carried out before the actual study.

This is done to:

• Identify potential problems with design or method

• Identify whether there is a chance of significant results being found

• So participants can suggest changes that could be made to improve the study

  • For example, participants not understanding the instructions, getting boring or guessing wjat the study is about

• The researcher can check that tasks aren’t too easy (ceiling effect) or too hard (floor effect)

confederates

Researchers can use other people to play a role in helping the investigation, they aren’t participants and they know what the investigation is about.

Also called a stooge, actor or pseudo-participant.

Studies involving confederates:

Bystander effect —

• Roughly hundreds of participants (people in the street)

• Actor pretending to be ill and participants in the street

• Being measured is the time taken for people to help the actor

Milgram study —

• 40 male participants

• Participant giving electric shocker (teacher) and actor/tape recording

• Being measured is how far people will hurt others because they are asked to by authority (and don’t see it as their responsibility

Asch study —

• 123 participants

• Participant in a group of actors 

• Being measured is how many people will follow what everyone else says and conform to the group when they know they are giving the wrong answer

aims and hypotheses

The aim is a statement of why the study is taking place and what it is trying to achieve.

For example, “The aim of the study was to investigate how caffeine affects exam performance”

The hypothesis states what you believe is the truth. It is a precise and testable prediction of the relationship between two variables.

For example, “Participants who are given caffeine before an exam will score 25% more points than participants who are not given caffeine”

operationalisation

A hypothesis needs to be operationalised so it is written in a testable form with precisely defined variables.

Simple hypothesis: “participants given caffeine will perform better than those that aren’t…”

Operationalised hypothesis: “16 year old students given coffee 15 minutes before exam will perform 50% better than those who aren’t…

Steps for operationalising:

• Identify participants clearly

• Give time frame

• Give statistical prediction (measurable unit)

• Specify type or measurement

types of hypotheses

Experimental and Null:

• Experimental hypothesis — prediction of what the researcher thinks will happen to the DV when IV changes

• Null hypothesis — prediction that states that the IV will have no effect on the DV and any observable different is due to chance

One-tailed and two-tailed:

• Directional / one-tailed hypothesis — predicts the direction of the difference between the two conditions after the experiment (specific)

  • For example, ‘drinking herbal tea improves memory’

• Non-directional / two-tailed hypothesis — predicts there will be a difference between two conditions but doesn’t state the direction (vague)

  • For example, ‘drinking herbal tea affects memory’

In exam:

• If stem mentions previously published research, used 1 tailed hypothesis

• If it states there is no previous research, used 2 tailed

• If it doesn’t mention anything about previous research then presume there is none and use 2 tailed hypothesis

variables

Independent variable — the variable that is manipulated or changed by researcher, the ‘cause’ in cause and effect

Dependant variable — the variable that the researcher is measuring, the ‘effect’ in cause and effect

The researcher wants to see how the IV affects the DV.

extraneous variables

Extraneous variables are the other variables that must be eliminated (removed all together) or controlled (made consistent across conditions) otherwise they will affect the DV and confound the results.

Types:

• Experimenter variables — to do with the researcher, e.g age/personality/ethnicity etc.

  • Controlled using standardised procedures (using same researcher for every condition so ppts have the same experience)

• Participant variables — to do with the ppts

  • Controlled by randomly allocating ppts to groups so differences cancel out

• Situational variables — to do with the situation, e.g time of day/temperature/materials/instructions

  • Controlled using standardised procedures and instructions

Demand characteristics — when ppts guess what study is about, includes:

• Trying to please researcher with results they want

• Trying to annoy researcher (screw you effect)

• Acting unnaturally out of nervousness or social desirability bias

—> controlled by single blind method, where ppts aren’t told which condition they’re in

Investigator effects — unconscious cues from investigator that encourage ppts to behave in a particular way, e.g body language/eye contact/tone etc.

—> controlled by double blind method, where ppts and researcher (research assistant) aren’t told condition

reliability and validity

• Reliability: consistency, if study can be repeated using same method and get same or similar results

• Validity: if test measures what it claims to be measuring

  • Internal validity — if we can be sure IV causes effects on DV or if confounding variables could be the reason, reducing EVs and using standardised procedures can improve internal validity

  • External validity — extent to which results can be generalised to other settings (ecological validity), people (population validity) and time (temporal validity), naturalistic settings improve external validity

sampling

Psychologists make predictions based on results from a sample because not everyone can take part.

• Target population — group that researcher wants to study and generalise results to

• Sample — small number of people that represent the target population

Sampling bias occurs when the sample is not representative of the rest of the population, having a larger sample can avoid this.

types of sampling

Random: each member has an equal chance of being selected

• Strength — in theory, it is best to generalise as it is unbiased

• Weakness — difficult to achieve with large target population, may still be unrepresentative as doesn’t ensure unbiased selection

Opportunity: asking whoever is around

• Strength — convenient

• Weakness — not likely to be representative as similar people hang around each other, people with certain characteristics are the most likely to take part, becomes a volunteer sample if people decline

Volunteer: people volunteer to take part

• Strength — convenient

• Weakness — only certain types of people will apply, demand characteristics (volunteers want to please)

Systematic: nth method

• Strength — higher chance of being generalisable as there is no researcher bias

• Weakness — may still be unrepresentative because doesn’t ensure unbiased selection, difficult with large population

Stratified: small scale reproduction of population, divided into important characteristics and randomly sampled in relative proportion

• Strength — representative

• Weakness — need details on target population (might not be available due to data protection laws), takes a lot of time

data

Types:

• Quantitative — numerical, objective, collected in experiment-based research

• Qualitative — non-numerical, subjective, collected in unstructured interviews, observations, case studies etc.

• Primary data — original data, collected specifically for research aim and hasn’t been previously published

• Secondary data — data collected for another research aim and has been previously published

• Meta analysis — combining the findings of several studies examining similar areas, allows for larger samples for easier generalisation

descriptive statistics

• Measures of central tendency: averages

• Measures of dispersion: how spread out scores are (variability)

• Percentages: rate within every 100

• Correlational data: positive, negative or no correlation coefficients

Measures of central tendency reduce data to a single representative value

• Mean — statistical average

  • Advantages: most sensitive (uses all scores), can be used with interval data

  • Disadvantages: can be distorted with extreme scores, might not be one of the actual values from the data

• Median — middle value

  • Advantages: unaffected by extreme values, can be used with ordinal data

  • Disadvantages: only takes into account one or two scores so not as sensitive and unrepresentative in a small set of data

• Mode — most often

  • Advantages: unaffected by extreme values, easy to calculate

  • Disadvantages: not useful in small sets or data or if there are too many modes (bimodal), doesn’t take into account all the scores

Measures of dispersion measure how spread out the data scores are

• Range — difference between lowest and highest value

  • Advantages: easy, takes into account extreme values

  • Disadvantages: distorted by anomalies, doesn’t show if data around mean is clustered or spread evenly

• Standard deviation — spread of scores from the mean

  • Advantages: more sensitive than range, more accurate as uses all the data

  • Disadvantages: difficult/complicated to work out

experiments

Types of experiments

Lab:

• Tightly controlled and ppts not in their natural environment

• The researcher deliberately manipulates the IV and controls extraneous variables

• Evaluations:

  • High degree of controls, leading greater accuracy and objectivity

  • Reliable, can be easily replicated

  • Low external validity because it is artificial and not like real life

  • Demand characteristics due to environment, unconsciously changing behaviour

Field:

• Conducted in real world environment

• Experimenter deliberately manipulates the IV and controls some extraneous variables

• Evaluations:

  • Greater external validity with more natural behaviour that can be easier to generalise

  • Less chance of demand characteristics

  • Difficult to establish cause and effect due to less control over extraneous variables

  • Unethical as ppts aren’t aware they’re being studied

Natural:

• IV varies naturally (experimenter has no control over it or extraneous variables)

• Evaluations:

  • High ecological validity

  • No demand characteristics, ppts may be unaware or researcher not present

  • Less control so low internal validity

  • Unethical

Quasi:

• IV is fixed naturally (e.g gender, age, ethnicity), meaning ppt has no choice of being randomly allocated to another group

• Regardless of natural, field, lab experiment, if IV is natural then it is now a quasi experiment

Experimental design

• Independent measures design — using different people in each condition

  • Quick and easy, ppts can be simultaneously tested

  • Avoids demand characteristics and order effects

  • More ppts needed for enough data

  • Ppt variability

• Repeated measures design — using the same people in each condition

  • Fewer ppts are required and more data

  • No ppt variability

  • Order effects (order they do can affect results, e.g bored or did worse)

  • Demand characteristics

  • Time consuming

• Matched pairs design — Using different people that are paired based on similar variables

  • No order effects or demand characteristics

  • Partially controlled ppt variability

  • Time consuming

  • Everyone is unique, no perfect matches

  • Data becomes useless if one twin drops out as both of the pair is needed

Counterbalancing (ABBA) cancels out differences found due to order effects, as if they become consistent across conditions then they are no longer extraneous.

observations

Observations are a non-experimental way to study behaviour by watching and recording it, most are naturalistic.

• Setting

  • Lab/controlled — artificial

  • Naturalistic — realistic

• Role of researcher

  • Participant — researcher is part of the group being studied

    • Richer insight and details but lacking in objectivity if researcher begins to associate w/ group

  • Non-participant — observer watches from a distance and isn’t involved in the group

    • Not affected by subjective thoughts but lacking in rich data

• Consent of ppts

  • Covert — ppts unaware they are being observed

    • Removes ppt reactivity which increases validity but ethically questionable

  • Overt — ppts are aware they are being observed

    • Ethical but ppt variables could confound study

Observations can suffer from observer bias since observers are interpreting data using their own feelings and expectations. Inter-observer reliability is when multiple observers decide to code behaviour in the same way before observation so it is more objective.

Structured observations used systematic sampling methods to record behaviour.

• Event sampling — counting the number of times a behaviour occurs

• Time sampling — recording behaving within a pre-established time frame

Observations have behavioural checklists with categories to check off. These should be mutually exclusive, precise and observable/measurable (not to do with emotions or thoughts etc.)

correlational studies

Correlations measure the relationship between 2 variables.

• Positive correlation — as one variable increases, so does the other

• Negative correlation — as one variable increases, the other decreases

Scatter grams are used to plot correlations.

Correlation coefficients:

-1 -0.7 -0.3 0 … 1

Perfect strong weak no perfect

negative negative negative correlation positive

Evaluation:

• Establishes relationship between 2 variables which can suggest areas for further researcher

• Statistical analysis can be done where experimental manipulation wouldn’t be ethical or practical

• Correlation =/= cause and effect, low internal validity

questionnaire

Questionnaires are a self-report research method where ppts give info about themselves. Researchers should try to select a large and representative sample for the research

Open (detailed, qualitative) and closed (fixed responses, quantitative) questions are both used.

Questionnaire construction:

• Questions to meet exact aim

• Length shouldn’t be too long

• Previous successful questionnaires can be used as examples for design

• Concise and easily understandable questions

• Pilot studies to get honest feedback

• Measure scales (likert scale) to asses psychological attitudes

Evaluation:

• Efficient to collect large amounts of data

• Qualitative and quantitative allows for detail and easy analysis

• Standardised questions, reliable

• Lack of investigator effects

• Social desirability and misunderstanding of questions

• Sampling bias due to kind of people who take the time to fill in surveys

• Not suitable for sensitive issues that require a researcher to be present to protect ppt

interviews

• Structured — prewritten questions, open and closed, interviewer doesn’t need much training

• Semi-structured — combination of structured and unstructured, produces quantitative and qualitative data

• Unstructured — not many predetermined questions and topics to explore (conversation-like), training needed

Evaluations:

• Detailed information (qualitative and quantitative)

• Conversation-style allows ppts to be honest and talk about sensitive issues

• Questions can be adapted or explained

• Structured/standardised interviews can be easy to replicate

• More time-consuming than questionnaires

• Difficult to analyse unstructured interview data

• Not suitable for all ppts who have to put thoughts and feelings into words

• Investigator effects (uncomfortable with certain researcher), demand characteristics and social desirability bias

ethical issues

• Deception

  • Not telling ppts the truth so they can’t alter behaviour

  • No informed consent and might make ppts untrustworhty of researchers in the future

    • Sometimes, informed consent isn’t possible, like: in research with children, deception is necessary and in field experiments

  • To resolve this issue:

    • Prior consent — before experiment, ppt agrees to be deceived but don’t know when or how

    • Presumptive consent — people similar to ppts asked if they’d be willing to take part, assumed that actual ppts would share their answers

    • Retrospective consent — consent gained after the experiment, also given the choice to withdraw results from study

• Briefing/debriefing

  • Ppts shouldn’t leave study with worse psychological state than they started

• Protection of participants

  • Distress is acceptable if it is the same than would be experienced in everyday life otherwise it should be stopped

  • Physical harm: stress and anxiety

  • Psychological harm: low self esteem, nightmares

• Right to withdraw

  • Ppts leaving could bias study (certain characteristics likely to stay)

  • Crying is used as a sign for children that they want to withdraw

• Confidentiality

  • Confidential data — tracked back to names

  • Anonymous data — collected without names

• Incentives