Current character dependency models (e.g., DELTA, p.19; DiversityDescriptions, p.322, XPER2 and CBIT Lucid3, p.21, see also Dallwitz 2006) focus on a special case of character-value de-pendency: certain values (states) of a categorical character determine whether another character is applicable or not (e.g., if “leaves absent” the leaf shape becomes inapplicable). In practice in-applicability based on categorical states is a frequent case and highly relevant for the efficiency of data recording and identification. However, it seems desirable to reserve the term “character dependency” for the general forms of character correlation and dependency. In contrast to the use in DELTA, the term “character applicability” is therefore preferred in this thesis for this special case.
The defining character in a character applicability relation is commonly called the controlling or parent character, the other the controlled, dependent, child or applicable/inapplicable char-acter. The values (categorical states) of the controlling character contained in a character appli-cability rule may be called controlling states, the remaining states defined in the character non-controlling states.
In current character applicability models, quantitative characters can be controlled but cannot be controlling. Examples where the applicability of characters depends on a specific value range of a continuously varying quantitative character are difficult to find. One example may be that for very small organisms or cell organelles (e.g., <0.5µm) light microscopic characters are no longer applicable (compare “Dependencies on circumstances of identification”, p.175). However, a de-pendency on counts of object parts is frequent. Supporting only categorical controlling characters forces the developers of a descriptive terminology to artificially split the character informing about the multiplicity of an object composition (see “Describing object multiplicity”, p.145) into two characters, one categorical for presence/absence, another for the number of objects if at least one is present. The latter is then usually controlled by the first one.
Character applicability rules may be expressed positively and negatively, called “inapplicable-if”
and “applicable-if” rules here. In the terminology used in this thesis, they may be formulated as:
■ “Inapplicable-if”: If only controlling states are present (i.e., “recorded” or “scored”) in a description, these make the controlled character inapplicable. The character remains applica-ble if either no state at all, or any non-controlling states are present.
■ “Applicable-if”: If any controlling state is present (i.e., “recorded” or “scored”) in a descrip-tion, these make the controlled character applicable. It is inapplicable if only non-controlling states are present.
Two special conditions are common to both forms:
■ If the controlling character is inapplicable (through another applicability rule, or through an explicit ‘inapplicable’ coding status value), the controlled character is always inapplicable as well (see “Cascading character applicability rules”, p.82).
■ If no data for the controlling character are present in a description (data completely missing, or only equivalent coding status; see p.74), the controlled character always remains applica-ble. In the case of an applicable-if-rule this definition is slightly unintuitive. However, sepa-rating the issue of default applicability from the form of the rule is a) generally desirable and b) required to achieve convertibility of the two forms (compare “Convertibility of applicabili-ty rules” below).
DELTA uses two directives (“Applicable/Inapplicable Characters”). The definitions in the user guide differ only in a single word (Dallwitz & al. 2000a): “This directive specifies the values of
‘controlling’ characters which make other ‘dependent’ characters [‘applicable’ in the definition of the Applicable Characters directive, ‘inapplicable’ in the definition of the “Inapplicable Char-acters” directive]. If, in a given item, a controlling character takes only values which make its dependent characters inapplicable, or if the controlling character itself is inapplicable, then the dependent characters must not be given any values (other than the pseudo-value ‘inapplicable’, which is redundant[…]).” Dallwitz (2006d) writes further: “If any of the recorded values of a given controlling character do not make its dependent characters inapplicable, then the depen-dent characters may be recorded. For example, ‘leaves present or absent’ allows other leaf char-acters to be recorded.”
These definitions are equivalent to those given above (the “values which make its dependent characters inapplicable” are the controlling states for inapplicable characters and the non-con-trolling states for applicable characters), except if the connon-con-trolling character contains no data in a description. The “applicable-if” definition given above assumes that the default of the controlled character is inapplicable, whereas in DELTA the controlled character remains applicable.
Before the convertibility of the two forms of character applicability rules can be discussed, the evaluation of the rules may need some clarifications:
A character may control multiple dependent characters (each with one or multiple controlling states) and a character may be controlled by multiple controlling characters (n:m relation). In the first case, applicability rules may be evaluated completely independently. For the second case, Dallwitz (2006d) defines the intended evaluation for DELTA as: “If a given dependent character
is dependent on more than one controlling character, then the dependent character can be re-corded only if allowed by all of its controlling attributes.” Note: in Dallwitz's terminology, an attribute is the set of all values for a single character in a single description.
Table 17. Examples of evaluating applicable/inapplicable-if rules involving multiple states and the same combination of controlling/controlled character.
States present for controlling character 1 in given description
Result of Inappli-cable-If rules 1.a →2 , 1.b →2
Result of Applicable-if rules 1.a →2 , 1.b →2
{1: a , b , c } 2 2
{1: a , b , c } 2 2
{1: a , b , c } 2 2
{1: a , b , c } 2 2
{1: a , b , c } 2 2
{1: a , b , c } 2 2
{1: a , b , c } 2 2
{1: a , b , c } 2 2 1
and indicate whether a state (indicated by lower case letters) in a description is scored or not; e.g., “1.a ” indi-cates that state ‘a’ of char. ‘1’ is present in a description. Applicable or inapplicable are represented by and .
1 Applicability in the case of no information for the controlling character: A literal interpretation of “applicable if state scored” would be “inapplicable”; however the DELTA interpretation is that the controlling character is unknown, the result of the rule unknown, and therefore the controlled character applicable (see p.80).
Furthermore, more than one state in a controlling character may control the same controlled character (Table 17). These states are to be evaluated together, splitting the states defined for a controlling character into a set of controlling states and a set of non-controlling states. Although the definitions given above are sufficient, a few points may be highlighted to avoid misinterpre-tations:
■ The set of controlling states is defined for each controlled character independently. Adding a rule “1.c →3 ” to the examples in Table 17 would not change the set of controlling states for character 2.
■ If multiple controlling states are present, any of these invokes the inapplicability (“1.a or 1.b”
in Table 17). It is not possible to establish a rule that is invoked only if multiple states (e.g.,
“1.a and 1.b” in Table 17) are present in a description.
■ The evaluation behavior within and between controlling characters is the opposite.
□ If within a single controlling character (and description) multiple states are present, any state that makes the controlled character applicable will result in the controlled character being applicable. In the case of an applicable-if-rule such a state is any controlling state, in the case of an inapplicable-if-rule such a state is any non-controlling state.
□ If within a single description multiple controlling characters are defined, any character that makes the controlled character inapplicable will result in the controlled character being in-applicable.
(For the difference within and between characters, compare also “Boolean operators between states of categorical characters”, p.95, and “Boolean operators between characters”, p.98.) Note that one may count all controlling states in a controlling character as a single rule, or one may count each controlling state as a rule of its own. In some scenarios of character evolution and federated (distributed) development of terminology the latter is desirable. However, this is a secondary matter of terminology and does not change the evaluation rules.