Random sampling picks who takes part in a study. Random assignment sorts those people into groups. Sampling supports generalization outward; assignment supports fair comparison inside.
- Random sampling, also called random selection, selects members of a population to form the sample.
- Random assignment places that sample into control and experimental groups.
- Sampling improves external validity, the power to generalize results.
- Assignment improves internal validity, the power to attribute results to the treatment.
- A study can use both, one, or neither.
- The sorting test is one question: does the sentence pick who takes part, or who goes where.
What random sampling is
Random sampling is a way to select members of a population for a study sample. Each member has a known or equal chance of selection. The goal is a sample that mirrors the population.
The Scribbr FAQ defines random sampling as selecting population members for the sample, in contrast with sorting that sample into groups (Scribbr FAQ). StatisticsSolutions gives the same split: sampling selects individuals from the population to take part (StatisticsSolutions).
Use random sampling when the goal is generalization. A survey of 1,000 voters works only when those voters represent the larger electorate. A study of school performance works only when the sampled students represent the school.
Random selection means the same as random sampling. The article uses sampling after this point, except when quoting the test.
Sampling does not guarantee a perfect mirror. Small samples still vary by chance. Practical limits also matter. A true random draw from a full list is often hard to reach. Researchers must not call a convenience sample random.
What random assignment is
Random assignment is a way to sort the sample into control and experimental groups by chance. Each participant has a known or equal chance of placement in any group. Studies that use simple random assignment are also called completely randomized designs.
Pritha Bhandari defines random assignment as placing sample members into treatment groups with randomization, with equal chance of control or experimental placement (Scribbr). StatisticsSolutions defines it as placing participants into groups such as experimental versus control (StatisticsSolutions).
Use random assignment in between-groups experiments. A clinical trial assigns patients to aspirin or placebo. A teaching study assigns students to a new method or a standard method. A therapy study assigns participants to different therapy types.
Assignment balances the groups at the start. Prior differences spread evenly by chance. The result then traces to the treatment, not to a prior gap. Bhandari warns that chance variation remains, especially with small samples, and that large samples reduce it (Scribbr).
The difference
Sampling comes first and answers who takes part. Assignment comes second and answers who goes where. Sampling looks outward to the population. Assignment looks inward to the groups.
| Feature | Random sampling | Random assignment |
|---|---|---|
| Question | Who takes part | Who goes where |
| Action | Select from population to sample | Sort sample into groups |
| Used in | Surveys, polls, observational work, experiments | Between-groups experiments only |
| Supports | External validity, generalization | Internal validity, fair comparison |
| Failure mode | Unrepresentative sample, results do not travel | Unequal groups, cause stays unclear |
GeeksForGeeks gives the same table in more detail, with purpose, application, process, outcome and focus rows (GeeksForGeeks). The version here keeps only the rows an exam requires.
Why the split matters
The split maps to two validities. Random sampling improves external validity. Random assignment improves internal validity. The Scribbr FAQ states both links in one sentence (Scribbr FAQ). Trochim states the same pair: selection relates to sampling and external validity, assignment relates to design and internal validity (Trochim).
External validity asks whether results generalize. A random sample of company employees supports claims about the whole company. A convenience sample of one break room does not.
Internal validity asks whether the treatment caused the outcome. Random assignment rules out systematic start differences. Without it, an alternative explanation survives: the groups already differed.
Read the wider validity family with reliability and validity. Reliability asks for consistency; validity asks for truth. Sampling and assignment are two tools that protect validity at different points.
How each is done
Both procedures start the same way. Give each person a unique number. Then randomize with a physical or computer method.
- 1.List every eligible person and assign a unique number.
- 2.For sampling, draw numbers by chance until the sample is full.
- 3.For assignment, draw numbers by chance to place each person in a group.
- 4.Record the method so another researcher can repeat it.
Bhandari lists four practical tools (Scribbr). They are a random number generator, a lottery draw from a hat, a coin flip for two groups, and a dice roll for three groups. Use a generator for large lists. Use a hat, coin, or dice for classroom demonstrations.
Do not use haphazard choice. Alternating sign-up order, picking postcodes, or assigning by recruitment site introduces bias. Bhandari shows the failure: recruits from gyms differ from recruits from cafes before treatment starts (Scribbr). Random tools remove that pattern.
Complex designs add one prior step. Group people into blocks by a shared trait, then randomize inside each block. A randomized block design does this to test whether the trait shapes outcomes. Name it only; exams rarely require the detail.
Can you have one without the other
Yes. Trochim shows all four combinations with plain cases (Trochim). The psychology versions follow the same logic.
| Sampling | Assignment | Example |
|---|---|---|
| Random, random | Draw 100 clients at random from 1,000, then assign 50 to a new program and 50 to control | Best case for both validities |
| Random, not random | Draw 100 students at random, then assign the first 50 to morning training and the rest to evening training | Generalizes, but group comparison is weak |
| Not random, random | Take the first 100 volunteers, then assign them by chance to sleep and no-sleep memory conditions | Fair comparison, but results may not generalize |
| Neither | Take two intact classes and let one teacher use the new method | Neither generalizes nor compares fairly |
Teachers use the third row often. A lab cannot sample the whole student body, but it can still randomize the volunteers it has. The result supports cause inside the lab and needs replication outside it.
Exam examples and the one-question test
AP psychology and undergraduate exams test the split with scenarios. Apply one question to each stem: does the procedure pick who takes part, or who goes where.
Example one: a researcher numbers all 8,000 employees and uses a generator to pick 300 for a survey. That is random sampling. It picks who takes part. It supports claims about the whole company. Scribbr uses this 8,000 to 300 employee case for sampling (Scribbr).
Example two: the same researcher takes those 300 and assigns each to a weekly team task or to no task by coin flip. That is random assignment. It sorts the sample into experimental and control groups. It supports claims about the task effect.
Example three: a memory study recruits volunteers from one course. It then assigns them by chance to a sleep group and a wake group before a recall test. Sampling is not random, but assignment is random. Cause is testable; generalization is limited.
The memory aid is two verbs. Sample selects; assign assigns. Selection points to the population; assignment points to the groups. Save the four terms as one set and drill them together: sampling, assignment, external validity, internal validity. To hold them apart, learn them against each other rather than one at a time, as memorise psychology terms recommends.
What random assignment does not fix
Random assignment does not make groups identical. It makes them comparable at the start, with residual chance differences. It does not remove extraneous variables entirely. It does not rescue a biased sample. A fair split of an unrepresentative group still generalizes poorly.
Some studies cannot use it. When the focus is a pre-existing trait, groups form by trait: men versus women, or people with versus without a condition. The trait is the independent variable, and all groups take the same test. Forcing chance placement would erase the research question.
Ethics also blocks it. A study cannot assign people to heavy drinking versus social drinking. When chance placement is impossible or unethical, researchers run a quasi-experiment on pre-existing groups and control other variables by design. Bhandari lists both exceptions: trait-based grouping and unethical assignment (Scribbr).
Do not confuse assignment with blinding. Assignment sorts people into groups. Blinding hides group membership from participants, staff, or analysts. A study can use both, but they solve different problems.
Frequently asked questions
What is the difference between random sampling and random assignment?
Random sampling selects who takes part from the population and supports generalization. Random assignment sorts those people into control and experimental groups and supports fair comparison. Sampling points outward; assignment points inward.
What is an example of random assignment?
Number each volunteer, then flip a coin for each number: heads to the experimental group, tails to the control group. For three groups, roll a dice: 1 or 2 to control, 3 or 4 to the first treatment, 5 or 6 to the second. Use a random number generator for long lists.
Does random assignment improve external validity?
No. Random assignment improves internal validity by balancing groups at the start. Random sampling improves external validity by building a representative sample. A study needs sampling for generalization and assignment for causation.
Can a study use random assignment without random sampling?
Yes. Many psychology experiments do exactly this. They recruit a convenience sample of volunteers, then assign those volunteers to conditions by chance. The cause is testable inside the study, but broad claims need replication with better samples.
What is random selection in psychology?
Random selection is another name for random sampling: selecting study members from the population by chance. Do not confuse it with random assignment, which sorts an existing sample into groups. Selection answers who takes part; assignment answers who goes where.
When do researchers not use random assignment?
Researchers skip it for pre-existing traits under study, such as age bands or clinical status. They also skip it when chance placement would be unethical. They then run quasi-experiments and control other variables by design.
Sources
- 1.Trochim, W. M. K. (n.d.). Random Selection and Assignment. The Research Methods Knowledge Base.
- 2.Bhandari, P. (2025). Random Assignment in Experiments: Introduction and Examples. Scribbr.
- 3.Scribbr FAQ. What is the difference between random assignment and random selection.
- 4.Statistics Solutions. Random Sampling vs. Random Assignment.
- 5.GeeksForGeeks (2025). Random Sampling vs Random Assignment.