Psychological Research Methods

Within Subjects (Repeated mesures)

• involves same participants taking part in two or more conditions of the same experiment.

Issues with within-subject design 

• Order effects 
• You will need to wait before drug is out of system
• Without controlling for carry over effects, the data could show there to be no differences in time estimation ability between the two conditions 
• Practice effects can occure

Between Subject

• Involves different participants taking part in different conditions 

Issues with between subject design

• Factors that influecne performance e.g. age, IQ and gender
• Consider matching participants on these factors
• Without matching it could lead to unrelated differences

Between Subjects 

Advantages - Comparison of very different groups and no order effects

Disadvantages - Groups may differ

Within Subjects 

Advantages - Meaningful comparison of data points because this design controls for individual differences 

Disadvantages - Order effects

§States that there are two statistical hypotheses
on offer

§The Null Hypothesis assumes that there
is no relationship or difference between the
two samples or variables under investigation

§The Alternative (Research, Experimental)
Hypothesis assumes that there is a
relationship or difference between the two
variables  

¨More typical of naturalistic studies ¨Often run in the field ¤Real-life settings, such as in communities ¨ ¨One example: ¤Measure maths anxiety in undergraduate psychology students and see whether there is any relationship between their maths anxiety scores and their end of year grades.

¨The basic laboratory experiment allows us to investigate causal relationships between variables ¤Does one variable affect another variable in a sequence of cause and effect? ¨Systematic variation of the level of the supposed causal variable whilst holding all other variables constant ¤See what effect the systematic variation has on the measured variable

¨Experimental manipulation ¨Standardisation of procedures ¤Controlling all variables other than the independent variable ¨Random assignment of participants to different conditions or orders of presentation

¨Keep all factors constant other than the variable being manipulated ¤Only the experimental manipulation should vary ¨Keep instructions, time of day, lighting, etc., the same between conditions ¤Eliminate variations in the behaviour of the experimenter ¨Avoid confounding of effects of the independent variable with other (extraneous) variables ¤Think of ‘confounding’ a bit like ‘confusing’

¨A control condition is identical to the other conditions, except that the proposed causal factor is not present ¤Acts as a baseline for comparison with the causal condition ¨There are ethical implications to consider when using a control group ¤Will they receive the drug treatment or educational intervention at a later date?

¨Depending on the design, there are two main methods ¤For between-subjects designs nAssignment to the control and experimental condition using a randomisation procedure ne.g., tossing a coin ¤For within-subjects designs nRandom assignment to the different orders of the conditions nVarying the order of conditions is call ‘counterbalancing’ (see slide 49) nHalf of participants gets condition A followed by condition B nOther half, condition B followed by condition A

¨‘Runs’ of the same condition may occur ¤So that the conditions are not completely similar- the run of participants may all be tested at one time of day and not spread over the whole day ¨The number of participants in each condition may not be equal ¤All of them could even be assigned to one condition! ¨Over the long-term, randomisation will equate the conditions but, in the short-term, chance may lead to differences between the conditions

¨One way round this problem is to use block randomisation ¤The first participant of every pair is randomly assigned to one of the two conditions, while the second participant of the pair is placed in the other condition nMatch the two participants on relevant variables ¤Ensures the same number of participants in each condition nIt is best to achieve equal numbers, since most statistical tests work best when there are equal numbers of participants in the different conditions

¨Where the experimenter does not have full control over the manipulation of the causal variable ¤Cannot randomly allocate participants to different conditions nGender nDepression nUnethical to randomly assign participants to ‘depressed’ and ‘control’ groups and then induce depression in the ‘depressed’ group! ¤Not easy to determine cause and effect as conclusively as with a true experiment

¨Another variable (or variables) can affect the relationship between the independent variable and the dependent variable ¤May account for some or all of the relationship between two other variables ¨Sometimes referred to as the ‘third variable problem’ ¤Although there may be many confounding variables! ¤ ¤ ¨Earlier example: body weight was considered a potential confounding variable

¨The procedures that we use to make observations should not introduce distortions ¨There are three types of distortion that could potentially affect the results of an experiment ¤Instrumentation ¤Observer bias ¤Sampling procedures

¨Any error/variation that is possible when measuring a variable
¨This error/variation is due to chance factors ¤i.e. it is random
¨There is no systematic bia ¤It affects the DV is both directions
¤Can be offset by taking large number of measurements and averaging them
¨It obscures the experimental effect ¤The effect of your IV on your DV

¨Biases or distorts the results of the experiment in a systematic way ¤The direction of the error will always be the same (systematic) nA constant error ¤Example: nAll the participants in the no alcohol condition were tested at 10am (fresh and rested) nAll participants in alcohol condition were tested in the evening (tired after a day at work/Uni) nThere is a systematic/constant bias in favour of the no alcohol condition ¤ ¨We must try to eliminate all sources of constant error from our experiments ¤Use the same testing time (or test an equal number participants in the different conditions at the two testing times)

¨We can remove constant error in two ways: ¤Direct control nRemoves the error nE.g. testing all participants at the same time ¤Counterbalancing or randomization nHowever, this does not eliminate the error, it just transforms it from constant error to random error nE.g. testing equal numbers of the ‘no alcohol’ and the ‘alcohol’ conditions at both testing times

¨There are many random errors that we cannot eliminate ¤An endless list of factors needing to be controlled ne.g. Health, IQ, personality, attitudes, experiences ¤Try to minimize error by choosing participants at random from the target population and allocate them randomly to different conditions nSo potential constant errors become random errors (the errors are averaged out across the conditions/groups)

¨Order effects ¤Differences between conditions are due to time, and not the experimental manipulation ¤Practice effect nImprovement in performance over time nE.g. taking the same experiment twice ¤Boredom effect nDecline in performance over time nE.g. doing the same task for 2 hours

¨Counterbalancing the order of presentation of conditions ¤Such that 50% of the participants do the experiment in the order Condition A followed by Condition B and the remaining 50% do the experiment in the order Condition B followed by Condition A ¨Randomisation ¤Random assignment of participants to do Condition A or Condition B first

¨Where the effect of an earlier condition affects performance of a subsequent condition, but not equally for different orders of presentation ¤Also known as the asymmetrical transfer effect ¨May occur if intervals between conditions are too close together ¤e.g., drug or alcohol studies ¤Have to wait for substance to clear the system

¨Two aspects to a drug ¤Its appearance ¤Its active ingredients ¨Participants have expectations about the effectiveness of a drug ¨The placebo effect is one in which the participants believe they have taken the active ingredient even though they haven’t ¤and produce similar results to the active treatment

¨Blind designs ¤Participant not told whether they are receiving the experimental or placebo treatment ¨Double-blind designs ¤Neither the participant nor the researcher know whether the experimental or placebo treatment has been administered ¤Ensures no unconscious clues being given to the participant by the researcher

¨Inaccuracy of observations ¤Are these errors consistently in favour of the experimenter’s hypothesis? ¨The experimenter expectancy effect ¤Unintentional influencing of participant’s behaviour towards that which the experimenter wants to see

¨The clues a participant picks up about a researcher’s expectations will determine a participant’s behaviour in the experimental setting (Orne, 1962) ¨To avoid this, we can either: ¤Develop a cover story to mask real intentions ¤Balance ethical considerations of lying to participants vs. gains in knowledge ¤Come up with a vague explanation ne.g., say we are investigating memory, rather than the specific research question, the effect of delay on serial recall