Plural for 'stratum'. Strata are groups or segments into which a population is divided, for the purpose of obtaining a representative sample. The material contained within each stratum is homogenous (similar) but that between strata is heterogeneous (different). The strata thus formed must be mutually exclusive, i.e. they must not overlap, and exhaustive, i.e. together they must cover all possibilities.
For instance, if a survey is to be conducted in a population and you want representation from all demographic groups in the population, you can consider each demographic group as a stratum. Then you can sample randomly from within each stratum (demographic group) to obtain a representative sample.
Ideally the sample from each stratum is chosen proportional to the size of the stratum, so that the composition of the sample reflects that of the population. In the above example, this means that if 80% of the population belong to one demographic group and two remaining groups cover 10% each, then a sample of size 100 should contain 80 people from the largest group and ten each from the smaller groups.
In situations where the population is made up of natural groups or strata, such as different demographic or age groups, the two sexes, different states/countries/industries, etc., using the method of stratification helps ensure every population group representation in the sample, where a simple random sample might fall short (in such a case a simple random sample might miss out a whole population group simply by random chance).