The evolution of heterostyly, a genetically controlled floral polymorphism, has been a hotspot of research since the 19th century. In recent years, studies on the molecular mechanism of distyly (the most common form of heterostyly) revealed an evolutionary convergence in genes for brassinosteroids (BR) degradation in different angiosperm groups. This floral polymorphism often exhibits considerable variability that some taxa have significant stylar dimorphism, but anther height differs less. This phenomenon has been termed "anomalous" distyly, which is usually regarded as a transitional stage in evolution. Compared to "typical" distyly, the genetic regulation of "anomalous" distyly is almost unknown, leaving a big gap in our understanding of this special floral adaptation strategy. Here we performed the first molecular-level study focusing on this floral polymorphism in Guettarda speciosa (Rubiaceae), a tropical tree with "anomalous" distyly. Comprehensive transcriptomic profiling was conducted to examine which genes and metabolic pathways were involved in the genetic control of style dimorphism and if they exhibit similar convergence with "typical" distylous species. "Brassinosteroid homeostasis" and "plant hormone signal transduction" was the most significantly enriched GO term and KEGG pathway in the comparisons between L- and S-morph styles, respectively. Interestingly, homologs of all the reported S-locus genes either showed very similar expressions between L- and S-morph styles or no hits were found in G. speciosa. BKI1, a negative regulator of brassinosteroid signaling directly repressing BRI1 signal transduction, was identified as a potential gene regulating style length, which significantly up-regulated in the styles of S-morph. These findings supported the hypothesis that style length in G. speciosa was regulated through a BR-related signaling network in which BKI1 may be one key gene. Our data suggested, in species with "anomalous" distyly, style length was regulated by gene differential expressions, instead of the "hemizygous" S-locus genes in "typical" distylous flowers such as Primula and Gelsemium, representing an "intermediate" stage in the evolution of distyly. Genome-level analysis and functional studies in more species with "typical" and "anomalous" distyly would further decipher this "most complex marriage arrangement" in angiosperms and improve our knowledge of floral evolution.