Breaking down the question
The question has two clear demands. First, explain the different types of non-probability sampling — this calls for accurate definitions of each technique. Second, bring out the conditions of their usage with appropriate examples — this calls for judgement about when and why each is chosen. Both halves must be answered fully; a catalogue of types without conditions of use will lose half the marks.
The pivot of the whole answer is the defining feature of non-probability sampling: units are selected not by random chance but by the researcher's judgement, so the probability of any unit being chosen is unknown. This is what distinguishes it from probability sampling and shapes when it is appropriate. The variables, sampling, reliability and validity theme grounds the treatment.
Choose vivid, sociologically apt examples — studying hidden or hard-to-reach populations, exploratory or qualitative work — so that the conditions of usage are concrete rather than abstract.
How to approach it
Begin by defining non-probability sampling against its probability counterpart, noting that it forgoes random selection and therefore statistical generalisation, but gains in feasibility and depth. State that it is the natural choice for qualitative, exploratory and hard-to-sample research.
Then take the main types in turn — convenience, purposive, quota, snowball, and volunteer or self-selection sampling — giving for each a definition, the conditions under which it is used, and a fitting example. Keep the structure parallel so the examiner can see each type treated on the same three points.
Close by weighing the trade-off: non-probability methods sacrifice representativeness for access, cost and depth, and are legitimate when those priorities dominate. This balanced verdict ties the answer together.
Model answer
Sampling is the process of selecting a subset of units from a population for study. In probability sampling every unit has a known, non-zero chance of selection, permitting statistical inference to the whole population. Non-probability sampling, by contrast, selects units by the researcher's judgement or by convenience, so the probability of selection is unknown and formal generalisation is not claimed. It is nonetheless indispensable where a sampling frame is unavailable, where the population is hidden or specialised, or where the aim is depth of understanding rather than statistical representativeness. Its principal types may be set out as follows.
Convenience sampling selects those units that are most readily available to the researcher. Its condition of use is speed, low cost and easy access, typically in pilot or exploratory studies where rigour is secondary to a quick preliminary reading. An example is a researcher surveying students in her own college canteen to pre-test a questionnaire. Its findings cannot be generalised, but it is useful for refining instruments and generating hypotheses.
Purposive or judgemental sampling selects units deliberately because they possess characteristics relevant to the research question. The condition of use is that the researcher already knows what kind of respondent will illuminate the problem, so expert judgement replaces randomness. It is common in qualitative and case-study work — for instance, choosing to interview experienced trade-union leaders to understand the dynamics of a strike. The strength is relevance; the risk is the researcher's own bias in deciding who counts as typical or informative.
Quota sampling divides the population into categories — by age, sex, class or religion — and fills a fixed quota within each, but the actual respondents are chosen non-randomly by the interviewer. Its condition of use is the wish to reflect the known composition of the population cheaply and quickly, as in much market and opinion research. An interviewer told to obtain fifty men and fifty women may simply approach whoever is convenient within each quota, which controls the profile but leaves selection within categories open to bias.
Snowball sampling begins with a few known members of a population who then refer the researcher to others, so the sample grows through social networks. Its condition of use is a hidden, stigmatised or hard-to-reach population for which no list exists — drug users, undocumented migrants, sex workers, or members of a secretive sect. A study of informal-sector migrant workers in a city, where each respondent introduces the next, is a typical example. It accesses populations otherwise invisible, though the network dependence can produce a clustered, unrepresentative sample.
Volunteer or self-selection sampling relies on respondents who come forward themselves, as in a magazine or online questionnaire. Its condition of use is breadth of reach at low cost, but volunteers tend to differ systematically from non-volunteers — often being more motivated or opinionated — so bias is severe. It suits exploratory canvassing of views rather than any claim about the population.
The choice among these techniques follows from the research situation. Where no sampling frame exists, snowball sampling opens a door; where the aim is theoretical insight from information-rich cases, purposive sampling is apt; where a rough demographic mirror is wanted quickly, quota sampling serves; where only a pilot is needed, convenience sampling suffices. In every case the gain is feasibility, economy and access to depth or to hidden groups; the cost is the loss of statistical representativeness and the intrusion of selection bias.
Non-probability sampling is therefore not an inferior substitute but a distinct toolkit, suited to qualitative, exploratory and hard-to-reach research where probability methods are impracticable or beside the point. The methodological maturity lies in matching the technique to the conditions of the enquiry, and in being candid about the limits on generalisation that the choice entails.
Examiner's perspective
Examiners look for both limbs of the question. Many candidates define the four or five types accurately but neglect the conditions of usage and examples, which is precisely where this question places its marks. Treating each type on the same three points — definition, condition of use, example — produces the tidy, comparable structure that reads well and is easy to reward.
The stronger scripts choose sociologically resonant examples, especially snowball sampling for hidden populations, and show why the technique fits the situation rather than merely naming it. A closing paragraph that frames the whole as a trade-off between representativeness and access signals conceptual command and lifts the answer into the top band. Confusing quota with stratified random sampling is a common error that examiners penalise, so the non-random selection within quotas should be stated explicitly.