While it is widely accepted that the semantic analysis of a stimulus can take place in an automatic fashion, it is typically assumed that non-automatic processes are required to process the relation of one stimulus relative to other stimuli. Nevertheless, there is evidence to support the idea that such relational stimulus processing can also take place under automaticity conditions. We examined this hypothesis further in four sequential priming experiments in which participants were asked to categorize target objects as larger or smaller than a reference object (i.e., a football or a car). Crucially, some primes were objects that were larger than the small reference object but smaller than the large reference object (e.g., a bike). Results showed that the impact of these primes upon target responding was dependent on the size of the reference object. When the size of the reference object was small, these primes facilitated responses towards large targets relative to small targets. Vice versa, when the size of the reference object was large, the same set of primes facilitated responses towards small targets relative to large targets. This result was obtained when the size of the reference object was manipulated block-wise (Experiments 1 and 3), trial-wise (Experiments 2 and 4), and even when the primes were presented near subjective recognition thresholds (Experiment 4). Taken together, our findings provide strong evidence for the hypothesis that complex relational stimulus processing can take place under automaticity conditions. A possible underlying mechanism is proposed.