observational techniques and design

what is an observation?

where a researcher observes and records ppt behaviour but there is no manipulation of variables and so no cause-and-effect relationship can be established so lower internal validity

can be used in experiment

advantages and disadvantages of observations?

• Capture people's actions - special insight into behaviour (people often do not do what they say they will)

  • See how people behave rather than how they say they behave - especially if they are unaware of being watched

• Allows us to study variables it would be unethical to manipulate

• Useful as pilot to generate hypothesis for future research

 

• Difficult to replicate

• Does not provide us with thoughts or feelings, only behaviour

• No manipulating variables - no cause and effect

• Observer effect - reduces internal validity

• Time consuming and requires careful preparation

• Observer bias - their interpretation of a situation may be affected by expectations (reduced by involving multiple researchers)

  • Ways to improve this are increasing number of observers

  • Recording the data so that it can be reviewed at a later date

structured observation advantages and disadvantages

predefined coding categories for behaviour - quantitative data

easier to record and analyse, easier to establish inter-rater reliability

reductionist, reduces validity, not much rich data obtained and open to observer bias

unstructured observations advantages and disadvantages

recording everything that is observed during observation - data will be qualitative

increased validity, applicable to many contexts, richer and more detailed data

harder to record, harder to establish reliability and more observer bias

controlled observations pros and cons

situation is contrived by researcher - in a lab like setting

higher internal validity due to less extraneous variables, more replicable as more standardised

low ecological validity, people do not act authentically and could be demand characteristics and social desirability bias

naturalistic observations pros and cons

observing ppts in their natural environments

high ecological validity and ppts will act more authentically

lower internal validity, more extraneous variables, low ecological validity if ppts realise they are being watched, less standardised and replicable

covert observations pros and cons

ppts don’t know they are being observed

higher internal validity as ppts act naturally, reduced ppt reactivity

ethical issues - no informed consent, invasion of privacy

overt observations pros and cons

ppts do know they are being observed and provide informed consent

reduces ethical issues

social desirability/demand characteristics and lower validity due to unnatural behaviour

participant observation pros and cons

researcher is involved with people they are observing

only way to observe behaviour, greater accuracy and detail, high external validity as increased insight into lives

hard to remain objective, influence behaviour which can reduce validity, ethical issues like deception, informed consent, invasion of privacy

non-participant observations pros and cons

researcher remains separate

easier to remain objective, no influence on behaviour, less ethical issues

less valuable insight, less detail and accuracy, observer bias, can’t observe certain behaviours

why must we pick a sampling method

• Sometimes an environment can be too busy to try and record everything that is happening

• Continuous recording - key in unstructured observations - all instances of target behaviour are recorded, may not be practical or feasible

• So decide how we will sample behaviour

  • event or time sampling

event sampling pros and cons

every occurrence of behaviour on predetermined checklist is observed and recorded within set time

less likely to miss events, increases validity and reliability

can be hard to record lots of behaviours happening at once, can miss events that are not coded for

time sampling pros and cons

behaviour on predetermined checklist is observed and recorded at specific time intervals

instantaneous scan - action performed at start of each interval recorded

predominant activity - researcher watches the entire time interval and notes the most predominant behaviour during the time

on-zero - researcher records whether a behaviour occurred or not in set time

more representative over time, easier and more reliable

can miss important events

what do behavioural categories need to be

clearly defined behaviours, observable and measurable, no dust-bin category, categories should not overlap

coding frames - allow for more specificity in categories (talking loudly or quietly)

• also to record severity and sub-types within categories

observer bias

observer’s personal qualities may influence their judgements of behaviours

reduce it with clear coding manual, extensive training for all observers and use a blind method when observing - don’t know aim

inter-rater reliability

researchers observe the same behaviour and code it in the same way

if low, suggests coding is vague/lacks validity

the observers have to agree first, not just loads of them

improve with pilot study, familiarise with categories, observe the same behaviour in pilot study, compare data and discuss differences in interpretations, calculate reliability by correlating the observations

agreement = 80% for high inter-rater reliability

ethical issues in observations

invasion of privacy, confidentiality respected, deception used, psychological harm may occur, lack of right to withdraw and informed consent, no debrief

public places - people should be expected to be seen by others anyway, you can observe people in areas they are already being observed

improving reliability and validity

reliability - clearly operationalised coding system, check inter-rater reliability, train researchers to use coding systems, conduct pilot study to check behavioural categories

validity - carry out covert observation so ppts don’t change behaviour, double blind to reduce observer bias, operationalised coding system

coder drift

• Coder drift results in a measurement error caused by gradual shifts in how observations get rated according to operational definitions, especially when behavioural codes are not clearly specified

• This type of error creeps in when coders fail to regularly review what precise observations constitute or do not constitute the behaviours being measured

prevent by operationalising codes, ongoing training, using reference videos, assess inter-rater reliability, adjust unclear codes