In some database systems it has become evident that transactions can consist of a succession of ‘tiers’ where different types of data are being accessed. Despite the different nature of the data accessed by each tier the transactions cannot be split up because of the interdependencies between the albeit different data of the individual tiers. However, the tiered nature of such transactions can offer greater scope for parallel execution and hence more efficient throughput of transactions without breaching data consistency. In this paper, we extend the standard transaction model by defining formally the notion of a tiered transaction. We give a concurrency control condition on tiered transactions being executed in parallel which is weaker than the standard serializability condition. We develop the theory in the context of the ‘syntactic’ methods of classical serializability theory.