Use of ontologies is a major trend of modern comparative, evolutionary and developmental biology (Dahdul et al., 2012; Mungall et al., 2012; Cooper et al., 2018; Walls et al., 2019). An ontology includes a set of standardized terms (each with a numerical code) describing features of structure and development as well as logical relationships between them. There are two major types of relationships: is_a (sepal is a kind of phyllome) and part_of (petiole is part of a leaf). Ontologies facilitate the assembly and analyses of large data sets. Larger data sets could be assembled from previously published subsets of data where the same characters are often coded under different names and in inconsistent ways. If characters of these disparate morphological data sets can be coded according to the conventions of plant ontology (such as those provided by Planteome, www.planteome.org), the codes may simplify merging various existing data sets. Even though the use of standardized ontologies is promising, it has some limitations as outlined below. Apart from assembling raw data sets, ontologies can be useful to link morphological and gene expression patterns, in large scale analyses of character evolution and potentially in resolving an ambitious task of reconstructing entire ancestral phenotypes (e.g., Sauquet et al., 2017), which are much more complex than merely combinations of many simple characters. Howard et al. (2021) highlighted the importance of the recently proposed method of reconstructing ancestral phenotypes using ontologies and stochastic mapping (Tarasov et al., 2019). The method constructs “gigantic multistate characters” by combining (amalgamating) individual traditional characters. Rather than producing complex matrices of character state transformation rates in such gigantic characters, Tarasov et al. (2019) employed a computationally more tractable approach of amalgamating stochastic maps of evolutionary histories of individual characters, that represent maps of character state changes along a tree topology. Ontologies should play key roles in such character amalgamation because they allow assembly of sets of related characters, for example, all characters related to a flower. The first step of character amalgamation is combining dependent characters. This critical step, however, must be performed manually. An example of dependent characters is perianth type (simple or double) and presence or absence of petal fusion. When perianth members are not differentiated into sepals and petals, the character of petal fusion makes no sense. Amalgamation of these two characters will create a more natural character with three states (simple perianth; double perianth with free petals; double perianth with fused petals). After fixing the problem of dependent characters, subsequent amalgamation of characters could be automated computationally using ontologies (Tarasov et al., 2019). The characters used in studies of morphological evolution are generally assumed to be based on homology (comparative homology, Vogt, 2017), but current plant ontologies are intended to be definition-based rather than homology-based (Dahdul et al., 2012; Cooper et al., 2018; Walls et al., 2019). Each term is introduced using a reference of a more general term and diagnostic characters of this particular term. The term petal can be defined as a phyllome (a general term) that is part of the corolla, defined as the inner perianth whorl, and is usually colored (diagnostic characters, see www.planteome.org). The use of clear definitions allows, in principle, the generation of stable and compatible data sets. Unfortunately, constructing completely definition-based and species-neutral (Mungall et al., 2012; Cooper et al., 2018; Walls et al., 2019) plant ontologies is technically impossible due to the complex nature of biological diversity. In the example above, petals may not be colored in some species (www.planteome.org). Hydrocotyle (Araliaceae) has a perianth composed of petals only (Erbar and Leins, 1985), while Sanguisorba (Rosaceae) has colored sepals and no petals (Wang et al., 2020). Only placement in a phylogenetic context (and not definitions) allows identification of their perianth organs as petals and sepals, respectively. In reality, existing plant ontologies combine features of the definition-based and homology-based approaches. For example, the ontology term spikelet is defined in the Planteome as “a reproductive shoot system that is the ultimate and congested inflorescence branch of the grasses”. This definition includes a reference to grasses (not other families with similar inflorescences). Why are one-flowered grass inflorescences (e.g., in Nardus) spikelets and not one-flowered capitula? Only because the experts working with grasses interpret spikelets of most grasses as homologous structures (but note the problem of pseudospikelets in bamboos, Kellogg, 2015). It was suggested that the grass spikelet shares some characteristics with a branch system and some with a flower because its developmental regulation is peculiar (Kellogg, 2000, 2015), but incorporating genetic data in definitions may cause a circular reasoning in subsequent comparative analyses of gene expression patterns. The effects of use of developmental genetic data in homology assertions (e.g., Mabee et al., 2020) should be carefully analyzed. The occurrence of the strong definition-based component in plant ontologies requires a transformation of data into a homology-based format before analyses of morphological evolution. For example, in an unbiased approach (e.g., Dahdul et al., 2012), Sanguisorba should be interpreted as having a simple perianth composed by tepals. At the step of manual transformation, a researcher should be able (but not obliged!) to re-score perianth members of Sanguisorba as sepals. This step is necessarily based on interpretations, and future research will show to what extent such transformations of data sets can be automated (Mabee et al., 2020). More than one interpretation is possible for many key structures. The grass palea is commonly interpreted as two fused outer whorl perianth members, but sometimes as a floral prophyll. The grass lemma is normally viewed as a flower-subtending bract, but may be a median outer whorl perianth member, or a novel organ that shares aspects of tepal (sepal) and bract identity (Kellogg, 2000, 2015). There are contrasting interpretations of the 3-staminate androecium found in most grasses (Rudall and Bateman, 2004). We argue that plant ontologies and associated software should be flexible enough to allow alternative homology assessments. A researcher would then be able to analyze the impact of different interpretations on large-scale analyses of morphological evolution, gene expression patterns and so on. Homology assessments of different characters in many cases should be coordinated with each other, which is a broad field for development of logical methods (Mabee et al., 2020). In the grass example, if the perianth is interpreted as two-whorled and the outer whorl as trimerous, then the inflorescence must be interpreted in such a way that flower-subtending bracts and floral prophylls are absent. Serial homology differs from homology in general in that states of the homologous features co-occur in the same individual organism (Nixon and Carpenter, 2012). The issue of serial homology is especially complex in organisms with open growth such as plants and some colonial animals as opposed to unitary organisms such as vertebrates or insects (Ochoterena et al., 2019). The terms “modular growth” (Hallé, 1986) and “modular organisms” (Notov, 2017) describing non-unitary organisms differ from the concept of modularity (Wagner, 1996) widely used in evolutionary developmental biology. A feature related to open growth is the occurrence of several hierarchical levels of serial homology (see Gatsuk, 1974). For example, monocot tepals are likely phyllomes (1st level). The outer and inner whorl tepals are usually dimorphic in commellinoid monocots (2nd level). Tepals may be dimorphic within a whorl (3rd level). If male and female flowers are present, their tepals should be separately described (4th level). Sometimes, flowers are dimorphic relative to their position in inflorescence rather than gender. Numbering these levels is arbitrary, but the enormous diversity and complexity of serial homology is obvious. A problem for digital approaches is that some levels of serial homology are not always sharply expressed. It is difficult to draw a boundary when serial homologues are different enough to be treated independently in plant ontologies. Some shoot systems include several distinct types of shoots and leaves, while these types are not expressed or only moderately expressed in other taxa. Different ontologies can be used depending on questions to be addressed in a given study. All analyses need not necessarily be based on inferred homology assessments. For example, the angiosperm egg nucleus is likely homologous to the non-angiosperm ventral canal cell nucleus (reviewed by Sokoloff and Remizowa, 2021). This interpretation is important in an evolutionary context, especially with respect to the enigmatic origin of angiosperms. However, the angiosperm egg cell is still an egg cell in the sense that its fertilization produces an embryo. It should be analyzed as such and compared to egg cells of other land plants in studies that focus on gene expression patterns governing fertilization. Archegonium necks of bryophytes and pteridophytes are not homologous to each other (see Renzaglia et al., 2000; Sokoloff and Remizowa, 2021), but studies related to sperm attraction and movement can use an ontology that includes a general term “archegonium neck”. Development and use of ontologies are important as a way of incorporating a holistic component in studies of plant evolution by linking individual characters to each other. We highlight the importance of comparative morphology in plant ontology research and suggest that evolutionary plant morphology should develop as a synthesis of computer algorithms and expert-based approaches where neither of the two sides predominate. The unstable number of levels of serial homology related to open growth is a plant-specific limitation in constructing a general ontology even for a fraction of land plants such as certain species-rich angiosperm family. This is one of the reasons for a scarcity of examples of the use of ontologies in large-scale studies of plant evolution. A solution that may make ontologies more useful is to allow more flexibility, including “modifying homology assumptions ‘on-the-fly’” (Mabee et al., 2020, p. 360). The user—a well-trained morphologist—should be able to select a subset of data and manually adjust it to keep more order and less chaos. This approach will allow use of ontologies for broad scale comparative and evolutionary analyses. The work is supported by the Russian Science Foundation (project 19-14-00055). We are grateful to Pamela Diggle, Richard Olmstead, and an anonymous reviewer for comments on the manuscript. Dmitry D. Sokoloff: Conceptualization (equal); Writing – original draft (equal). Margarita Remizowa: Conceptualization (equal); Writing – review & editing (equal). D.D.S. and M.V.R. wrote the manuscript.