Examination of three variable concord phenomena in spoken Portuguese of Rio de Janeiro (noun phrase, subject/predicate adjective, subject/verb) demonstrates that the most consistent and strongest variable constraint on the use of concord is the discourse level serial (or parallel processing) effect, that is, a preference for similar forms to co-occur in sections of discourse, even if this results in redundancy or lack of economy. Given the strength of this constraint, we separated the data into two subsets such that one contains only serial occurrences, where parallel processing can operate, whereas the other contains only isolated or first occurrences, where there is no preceding form to trigger the parallel effect. In the serial set, the parallel effect is so strong that other nonsocial constraints are not selected. External forces, such as the normative pressure of the school system toward fully redundant marking, have no significant effect on the serial tokens, although education does correlate with increased use of concord in isolated/first tokens. Thus, in the serial context the linguistic system is so highly constrained that external, and even internal, influences are blocked. We conclude that strong constraints should be identified, and that data in which they are operative should be analyzed separately from data in which they are not present in order to understand the interplay of the diverse forces in the linguistic system.