Investigation Design

Investigation Design
A reasonably precise statement of why a study is taking place. It should include what’s being
studied and what the study is trying to achieve.
A hypothesis is much more precise than an aim and predicts what’s expected to happen. They
are testable statements.
Experimental hypothesis. Used for experiments only. This predicts significant
differences in the DV as a result of manipulation of the IV. Any difference or effect
found is not due to chance. Alternative hypothesis is used for all other research
a) Directional/one-tailed hypothesis. This states the direction of the results. It is
called directional because it states one direction in which the results can go.
b) Non-directional/two-tailed. This states there will be a difference, but doesn’t state
the direction of the results.
Null hypothesis. It predicts that the IV will not affect the DV, the results will simply be
due to chance
Repeated measures design. The same participants are tested in the 2 (or more) conditions.
No individual differences between the 2
Fewer participants are needed (half as
Order effects. Can be controlled by
counterbalancing “ABBA”
Lost participants (may drop out)
Demand characteristics – more likely to
guess the purpose of the study
Takes more time. May need a gap between
Independent groups design. Different participants are used in each of the conditions.
Participants are often randomly allocated to each condition.
No order effects
Less chance of demand characteristics
Need more participants
Any differences between the groups may be
due to individual differences not the
manipulation of the IV. Random allocation
would reduce this problem.
Time saved
Matched pairs design. Different but similar participant are used in each of the conditions.
Participants are matched across groups on any characteristics judged to be important for that
particular study e.g. age, gender, ethnicity.
See independent group design
Less differences between the groups
See independent group design
Matching is difficult. It is impossible to
match all variables between participants.
The variable missed might be vital.
( Identical twins would be best)
Time consuming. Takes a long time to
accurately match participants on all
Design of naturalistic observations
Data can be collected by using visual recordings, audio recordings (usually analysed later) or
on the spot note taking using previously agreed rating scales or coding categories. One
difficulty with naturalistic observations involves the development and use of appropriate
behavioural categories.
Behavioural categories. – Observers have to agree on a grid or coding sheet on which to
record the behaviour they wish to study. It is often easier to code or rate the behaviour e.g. T
= talking. M-P = using mobile phone. Or a rating scale 1-5 e.g. on a scale of “safe driving”.
One of the main problems is to achieve standardisation between different observers and a lot
of training is required. Checks that all observers are coding behaviour in the same way ensure
inter-observer reliability. One way to assess this is to conduct a correlation of all the
observers scores. High correlation indicates reliability ( may not be correct but is consist as a
Design of questionnaire
There are a number of essential factors in questionnaire design.
Aims – Be very clear about the aim of the research, only ask questions that address
these aims
Length – The longer the questionnaire the less likely it will be completed. Make it
short and to the point.
Advice. Examples of questionnaires that have proved successful in the past should be
used as a basis for the questionnaire design
Statistical analysis – this should be considered at the design stage
Presentation – looks matter
Question order – useful to start with some simple factual biographical questions
before moving on to more probing questions. Usually best to put essential questions
in first half as many are sent back unfinished. Keep it interesting
Question formulation – questions should be simple, to the point and easily
understood. Avoid ambiguity, complicated terms, double barrelled questions etc.
Incentives – offering an incentive to complete a questionnaire can help with return
rates. It should also be easy to return e.g. a pre paid envelope.
Pilot study – A test of the questionnaire should be done on people who can provide
detailed and honest feedback.
Measurement scales. When using scales e.g. 1 strongly agree 2 Agree 3 undecided 4
disagree 5 strongly disagree – many will choose the “undecided” but unclear whether
this is because they have no opinion or cannot decide between their attitudes. Likert
scale possibly best known attitude scale.
Design of interviews
It is important to consider who will be the interviewer.
Gender and age of the interviewer affect the answers of respondents especially if the
topic is of a sensitive sexual nature.
Ethnicity – some interviewers have more difficulty interviewing people drawn from a
different culture and have less rapport with the participant
Personal characteristics – some people are easier to get on with, interviewers may
adopt different roles. Use of language, accent and appearance can also affect how
someone comes across to the interviewee.
Interviewers need training. Need to be able to listen. Non-verbal communication is also
important. Need to be able to relax interviewee. The more difficult and probing questions are
usually best left to the end
Operationalisation of variables
This means being able to define variables simply and easily in order to manipulate them (IV)
and measure them (DV). Sometimes this can be difficult to do and if the issue is complex then
when you operationalise you may only measure one aspect of the variable. If the IV and DV
are not operationalised accurately and objectively the results may not be reliable or valid.
Pilot studies.
These are small scale ‘practice’ investigations where all aspects of the research can be
checked and any changes can then be made.
Control of extraneous variables
Extraneous variables must be carefully and systematically controlled. Researchers should
Participant variables – age, intelligence, personality etc. should be controlled across
the different groups taking part.
Situational variables - the experimental setting and surrounding environment must be
Experimenter variables – the personality, appearance and conduct of the
experimenter should be controlled
If the extraneous variables are not controlled they can become confounding variables which
minimise the value of any results.
Reliability and validity
Reliability – If results are reliable they are said to be consistent i.e. you would get similar
results if you repeated the study. Results can be reliable but not accurate (faulty equipment)
or valid
Validity – Measure what they’re supposed to be measuring and are accurate.
Internal reliability – is a test consistent within itself.
External reliability – is test reliable over time. Check with test retest method.
Internal validity – are results valid and directly attributed to the manipulation of the
IV (not affected by confounding variables). To be internally valid there should be : no
investigator effects, no demand characteristics, standardised instructions should be
used as well as random sampling.
External (ecological) validity. Can the results be generalised to the wider population or
to different settings or different historical times.
ETHICS – Should follow The British Psychological Society’s Code of Ethics
Informed consent. Where possible participants should be informed of the objectives
and if participant parental consent should also be given
Avoidance of deception – deception should be avoided where possible this includes
withholding information or misleading participants especially if participants likely to
object or show unease in a debrief. To deal with this issue can use a) presumptive
consent, b) prior general consent or c) retrospective consent.
If deception is used, participants should be told immediately afterwards and given chance to
withdraw their data. The investigator should ensure that alternative procedures that do not
involve deception are unavailable.
Adequate briefing/debriefing – all relevant details of the study should be explained to
participants before and after the study. Debrief is especially important if deception
has been used. Participants should ;leave the study feeling the same or better about
themselves as when they started.
Protection of participants – must protect participants from physical and mental harm.
Normally the risk of harm must be no greater than in ordinary life
Right to withdraw – participants should always be aware they can leave the study at
any time regardless of any payment and withdraw their data at any point. Particular
issue with observations.
Confidentiality – information should be kept confidential unless an agreement has
been made in advance. Use numbers instead of names (confidential – can be traced
back to names, anonymous – cannot be traced back)
Observations – should only be made in public places where people might expect to be
seen by strangers.
Giving advice – if evidence of a psychological or physical problem of which a
participant is apparently unaware is obtained the investigator has a responsibility to
inform the participant if he/she believes that by not doing so the participant’s future
well being may be endangered.
Colleagues – investigators share responsibility.
Before starting research psychologists should, seek peer guidance, consult likely participants,
follow BPS Code of Ethics, consider alternative methods, establish a cost benefit analysis of
short and long term consequences, assume responsibility and gain consent from any ethical
committee. If negative effects are found during the research it should be stopped
immediately and every effort should be made to correct adverse consequences.
Research Method
Lab experiments
Field/natural experiments
Ethical Issue
 Participants feel pressure to act in a particular
 Reluctance to exercise right to withdraw
 Experimental situation can be stressful
 Informed consent difficult to get
 May be unaware of right to withdraw
 Debriefing is difficult
 If unaware of observation, issues of informed
consent, confidentiality and invasion of privacy
Correlational analysis
Case studies
Correlations may be interpreted incorrectly by
the public
Confidentiality must be maintained
Right to withdraw information on embarrassing
Issues of confidentiality and invasion of privacy
The sample should be representative of the target population. This is the group of individuals
the researcher is interested in. Generally the larger the sample the better.
Random sampling. Every member of the target population has an equal chance of
being selected. Use computer generated random lists or place all names in target in a
hat and pick out required number (still a chance sample could end up biased) but is
more likely to be representative and therefore results can be generalised. It can be
difficult to get details of the whole target population.
Opportunity sampling – select participants who are readily available and willing to
take part
Volunteer sampling – involves people volunteering to take part
Advantages of opportunity and volunteer sampling – easy, practical, cheap.
Disadvantages of opportunity and volunteer sampling – Sample likely to be biased so
findings may not be generalised to target population.
Demand characteristics
Guessing the purpose of the research and trying to please the researcher by giving
‘right’ results
Guessing the purpose of the research and trying to annoy the researcher by giving
‘wrong’ results the ‘screw you’ effect
Acting unnaturally e.g. out of nervousness
Acting unnaturally in order to look good –social desirability bias
Single blind techniques reduce demand characteristics
Investigator effects
Physical characteristics of researcher may influence results
Other characteristics such as accent, tone of voice, eye-contact, smiling etc. can
influence results
Investigator may be biased in their interpretation of the data
Double blind techniques reduce investigator effects