Floristic identification of species belong to the genus Dictyota J. V. Lamouroux (Phaeophyceae, Dictyotales), it has been a controversial issue among the community of Dictyotales specialists for at least the past two decades [1] - [3] . This scenario was the consequence of that nineteenth-century phycological school which saw the least minimal morphological difference as a diagnostic feature (the classical algorithm for that is: if one cell, then species A; if two, then species B), which in turn was the consequence of a very low morphological variability and phenotypic plasticity, both intra and inter species. Today, 77 specific names are recognized within the genus worldwide [4] . For the American Atlantic, for example, we recognize 15 species [5] , and 11 for the Mexican Caribbean alone [6] - [9] . Fortunately, methods of modern molecular phylogenetics have allowed us to solve the problems resulting from the lack of reliable morphological characters. As an instance for genus Dictyota, molecular characters have been used in order to redefine intraspecific and interspecific taxonomy [10] - [16] . Recently, several authors studied the taxonomic relationships of Dictyota species at the interspecific level, combining mor- phological and molecular data [3] [17] - [20] . As a result of these studies, new species had been described within Dictyota [3] [18] [20] . Because of the above, we started a molecular characterization of genus Dictyota along the Mexican coasts of the Gulf of Mexico and the Mexican Caribbean. During the research, however, it was found that two particular sample collections did not match the sequences published for a total of 39 species of Dictyota. A high genetic divergence, together with morphological analyses, convinced us of the presence of two new species, which here we formally described.

The specimens studied in the present study were collected at Punta Brava and Puerto Morelos, Quintana Roo, in March 2012, by J.G. Lozano-Orozco and A. Sentíes. Voucher samples were fixed in 4% formalin solution for morphological analysis.

The specimens were examined using a Leica stereomicroscope (Leica MZ 12.5, Wetzlar, Germany). Transverse hand sections in the apical, middle and basal regions of the thallus were made with a stainless steel razor blade under a dissecting scope. Photomicrographs were taken with an Olympus Vg-160 digital camera (Tokyo, Japan) coupled to a Nikon Eclipse E200 microscope (Tokyo, Japan). Voucher specimens were deposited at UAMIZ Herbarium (UAMIZ1231, UAMIZ1232, UAMIZ1233 and UAMIZ1234).

Total DNA was extracted from silica gel-preserved material using a Dneasy Mini Kit (Qiagen, Valencia, CA, USA) following the manufacturers’ instructions. A total of 956 bp from the psbA gene were amplified using the psbA F and psbA R primers [16] , and 672 bp were amplified for cox1 with Gaz1R and Gaz1F primers (adapted from [21] ), using the AmpliTaq kit of the Applied Biosystems (Lincoln Ventre Drive Foster City, CA, USA). All PCR products were checked for their correct length by electrophoresis on 1% agarose gel. The PCR products were sent to Macrogen (Gasan-dong, Seoul, Korea) to be purified and sequenced. Alignments for psbA and cox1 sequences were constructed using MEGA version 5 [22] . Collection sites and GenBank Accession Numbers of the specimens used in the phylogenetic analysis are provided in Table 1.

Phylogenetic relationships were inferred using a concatenated matrix (1703 nt) of two genes from 39 taxa (psbA = 1031 bp; cox1 = 672 bp; Table 1). Canistrocarpus crispatus (J. V. Lamouroux) De Paula & De Clerck, Dictyopteris polypodioides (A. P. De Candolle) J. V. Lamouroux, Rugulopteryx okamurae (E. Y. Dawson) I. K. Hwang, W. J. Lee & H. S. Kim and Scoresbyella profunda Womersley were included as outgroup taxa based on previous results obtained by [23] . The intraspecific and interspecific divergence values obtained for psbA and cox1 were calculated using uncorrected “p” distances implemented in MEGA version 5 [22] . The Bayesian phylogenetic analysis was inferred with MrBayes v.3.0 beta 4 software [24] using the general time-reversible model of nucleotide substitution with invariant sites and Gamma-distributed rates for the variables sites (GTR + I + G). This model was selected based on Maximum Likelihood (ML) ratio test implemented by the software ModelTest 0.1.1 [25] . For the Bayesian analysis, five chains of the Markov chain Monte Carlo (one hot and four cold) were run, sampling one tree every 1000 generations for 5 × 10⁶ generations starting with a random tree. Maximum likelihood analysis was performed using TOPALI v2 software [26] with the GTR + I + G model. The ML bootstrap analyses were conducted with 100 replicates. Support values [27] were computed, as implemented in TOPALI v2.