More explicit approaches usually work at the lowest level of parallel control, providing sets of basic parallelism primitives that could then be exploited to build more complex structures such as evaluation strategies. The approach is typified by MultiLisp [Halstead, 1985] or Mul-T [Kranz et al., 1989] which provide explicit futures as the basic parallel control mechanism. Futures are similar to GpH pars.
At a slightly higher level, Jones and Hudak have worked on commutative Monads [Jones and Hudak, 1993], that allow operations such as process creation (called fork) to be captured within a standard state-transforming monad. While this approach provides the essential building blocks needed to support evaluation strategies, it has the disadvantage of raising all parallel operations to the monad level, thus preventing the clean separation of algorithm and behaviour that is observed with either evaluation strategies or first-class schedules.