As a tribute to A.C. Martin's classic work on embryos in seeds, we have attempted to gain a better understanding of the peripheral embryo, which puzzled Martin. The peripheral embryo is strongly curved and in contact with the inner surface of the seed coat, and Martin placed it at the base of his family tree of seed phylogeny and called it a ‘phylogenetic orphan’. We evaluated ovule/seed development, kind of embryo and occurrence of perisperm in families with and without a peripheral embryo. All families with a peripheral embryo occur in the Caryophyllales. Seeds with a peripheral embryo have a low cotyledon width:radicle width ratio that coincides with Martin's (full-sized) linear embryo. The peripheral embryo develops in campylotropous and/or amphitropous ovules and is pushed to the side of the seed as the perisperm develops. Linear-full embryos and perisperm are widely distributed across extant angiosperms but are rarely found together, except in core Caryophyllales. The non-core Caryophyllales with endosperm and various kinds of embryos, including the linear-full, diverged before the core Caryophyllales. Thus, the ancestral linear-full embryo appears to have been retained when the core lineage developed campylotropous and/or amphitropous ovules and perisperm. Seeds with a peripheral embryo merit a position on Martin's family tree; however, the position should be a side branch (‘orphan’) slightly above (more advanced than) his linear embryo and not at the base. We conclude that Martin had great insight into the relationships between the kinds of embryos and rightly questioned the position of the peripheral embryo.