What, if anything, guides the evolution of Earth surface systems (ESS)? Goal functions and attractors are attempts to explain or predict where landscape evolution is headed, and why. Goal functions and attractors may be roughly grouped into three categories—single path, single outcome; alternative stable states; and multiple path, multiple outcomes. The latter includes systems characterized by dynamical instability and deterministic chaos, where small disturbances or variations in initial conditions become magnified (amplification selection). A classic example—the Lorenz model of atmospheric convection—is discussed. Traditional theories of succession or progressive development to a climax form are single path/outcome attractors common in geology, geomorphology, pedology, and ecology. Various forms of selection play a key role in each of these. Steady-state equilibrium and self-organized criticality concepts are also single path/outcome constructs, sometimes associated with purported goal functions, but that actually represent emergent selection. The example of river longitudinal profiles is presented. Power law distributions are often presented as attractor states, or as evidence of attractors. However, power law distributions alone are problematic evidence, as many phenomena can produce these. Alternative stable state concepts involve multiple (usually two) stable attractor states separated by instability, typically in threshold-dominated systems. Goal functions that are often described as ESS seeking a particular state can be explained via emergent selection. As this does not imply any intentionality of ESS, the emergent selection explanation is preferred over goal functions by Ockham’s Razor.