This glossary of Unified Modeling Language terms covers all versions of UML. Individual entries will point out any distinctions that exist between versions.

• abstract - An indicator applied to a classifier (e.g., actor, class, use case) or to some features of a classifier (e.g., a class's operations) showing that the feature is incomplete and is not intended to be instantiated.
• abstract class - A class that does not provide a complete declaration, perhaps because it has no implementation method identified for an operation. By declaring a class as abstract, one intends to prohibit direct instantiation of the class. An abstract class cannot directly instantiate objects; it must be inherited from before it can be used.
• abstract data type
• abstract operation - Unlike attributes, class operations can be abstract, meaning that there is no provided implementation. Generally, a class containing an abstract operation should be marked as an abstract class. An Operation that must have a method supplied in some specialized Class before it can be used.
• abstraction is the process of picking out common features and deriving essential characteristics from objects and procedure entities that distinguish it from other kinds of entities.
• action - An action is the fundamental unit of behavior specification and represents some transformation or processing in the modeled system, such as invoking a method of a class or a sub activity
• action sequence -
• action state -
• action steps -
• activation - the time during which an object has a method executing. It is often indicated by a thin box or bar superimposed on the Object's lifeline in a Sequence Diagram
• Activity diagram - a diagram that describes procedural logic, business process and work flow supporting parallelism
• active class - a class defining active objects
• active object - an object running under its own thread
• activity - carrying out behaviour in a State machine diagram
  • Do - a type of Activity which may be interrupted, as opposed to normal Activities which may not be interrupted
  • Internal - an Activity that is executed within a State
    • Entry - an Activity that is executed when a State is entered
    • Exit - an Activity that is executed when a State is exited
• activity final - the end point of an activity diagram
• actor - a role that a user takes when invoking a use case. Also see actor modeling.
• aggregation - a special type of association used to represent a stronger relationship between two classes than a regular association; typically read as "owns a", as in, "Class A owns a Class B". A hierarchy of classes where the child objects continue to exist even if the parent object is destroyed; see 'composition'
• artifact - items that model physical pieces of information in your system, such as a user's manual, training material, or password file
• association - a relationship with 2 or more ends, where each end is on a class (or other classifier). Each end may have a Role, Multiplicity, and be Navigable
• association class - a class that describes an association.
• asynchronous -
• attribute - a significant piece of data owned by a Class, often containing values describing each instance of the class. Besides the attribute name and a slot for the attribute value, an attribute may have specified Visibility, Type, Multiplicity, Default value, and Property-string.

• behavior - The observable effects of an operation or event, including its results.
• behavior diagram

• call -
• cardinality - the current number of occurrences of a Property. The cardinality must be a value that is allowed by the multuplicity
• child -
• class - the primary declarative construct of object-oriented programming; a cohesive unit of Attributes and Operations; a compile-time template for an Object
• class diagram - a type of static structure diagram that describes the structure of a system by showing the system's classes, their attributes, and the relationships between the classes.
• classifier - a category of UML elements that have some common features, such as attributes or methods.
• client -
• collaborate -
• communication diagram
• component -
• component diagram
• composite structure diagram
• composition - a specific type of relationship describing how one Object is composed of another Object; compare with Aggregation
• condition -
• connection -
• constraint - natural language, programming language or Object Constraint Language boolean condition which may not be false if a Class is to be considered valid

• decision - a point in an Activity diagram where a Flow splits into several, mutually exclusive, Guarded flows. A Merge marks the end of the optional behaviour started by the Decision
• dependency - a dependency exists between two defined elements if a change to the definition of one would result in a change to the other. In UML this is indicated by a line pointing from the dependent to the independent element. * deployment diagram
• derived property - a property that can be calculated or inferred from other properties
• diagram - a visual representation of a subset of features of a UML Model
• directed relationship -
• domain -

• edge - a synonym for Flow
• entity -
• enumeration - a set of constant values for a new data type
• event - when it occurs on an Object it may cause a Transition in a State machine diagram
• expansion region - a set of Actions in an Activity diagram that occur once for each of a collection of input Tokens to the Expansion Region
• extend -
• exception -

• file -
• final state - the state at which an object ceases to exist
• flow - a navigational connection between two Actions
• flow Final - the point at which a Flow ends without ending the complete Activity
• fork - a point in an Activity diagram where a Flow of logic splits into several concurrent Flows (Threads)
• found Message - starting point for a Sequence diagram

• generalization - a relationship between a specific classifier (typically a class) to a more general classifier asserting that the general classifer contains common features among both the specific classifier and the general classifier. Features include, for example, properties, and constraints. The use of generalizatiion is often logically restricted to cases where the specific classifier is a "kind-of" or "sort-of" the general classifier: for example, a Boxer is a "kind-of" Dog. When the classifiers involved are software engineering classes, generalization usually involves reusing code; it is often implemented using inheritance, where the more specific code reuses the more general code.
• generalization tree - Several specialized classifiers may point to the same general classifier, forming a generalization tree, where the general classifer contains common features shared by all the specialized versions. As generalization is a relationship, it is possible for a classifer to participate in several generalizations, often being on the specific end or on the general end, forming a directed acyclic graph (DAG) (i.e., no loops).
• generic -
• guard - a Boolean test that must be satisfied for a Flow of an activity diagram or a Transition of a state machine diagram to be allowed to start

• hierarchy -
• history pseudostate - points to the initial state of an object where no previous state history was saved

• inheritance - where a new more specific Class derives part of its definition from an existing more general Class
• initial node - the start point of an Activity diagram
• initial pseudostate - points to the initial State of an Object
• interaction diagram
• interaction overview diagram
• Interaction Frame - a section of a Sequence diagram, divided into fragments, which is subject to an algorithmic Operator such as iteration, parallelism or optionality.
• interface - a defined communication boundary.

• join - a point in an activity diagram where several concurrent flows (threads) synchronize, waiting until all are complete before continuing with a single flow

• lifeline - indicates a participating Object or Part in a Sequence diagram. The Lifeline may show activation, Object creation, and Object deletion.
• link -

• merge - a point in an Activity diagram marking the end of the optional behaviour started by a Decision
• message - a signal from one object (or similar entity) to another, often with parameters. Often implemented as a call to a Method, including the Constructor and Destructor, of an Object in a Sequence diagram.
• metadata -
• meta-model -
• meta-modeling -
• meta-modeling technique -
• meta-object facility -
• method -
• modeling -
  • domain - the representation of real world conceptual entities
  • design - the representation of software Classes and Interfaces
  • dynamic - use of Interaction diagrams to describe collaborations and behaviour
• model -
• model-driven architecture (MDA) -
• multiplicity - a specification of the number of possible occurrences of a property, or the number of allowable elements that may participate in a given relationship. In UML 1.x, it was also possible to have a discrete list of values, but this was eliminated in UML 2.0.
  • mandatory - A required multiplicity, the lower bound is at least one, usually 1..1 or 1
  • optional - The lower bound is at most zero, usually, 0..1
  • many - A multiplicity with no upper limit, either 0..* or *

• namespace - a context in which an identifier exists.
• navigable - the ability for objects of a Class at one end of an Association to retrieve Objects from the other end. Associations need not be navigable.
• node -
• note -

• object - a runtime instance of a Class. Objects are rarely shown on diagrams (because there are usually too many) unless they are used to illustrate some scenario, test, etc. Such Objects are often shown with the Attributes of the Class populated with sample data
• Object Constraint Language (OCL) - a declarative language for describing rules and constraints that apply to UML models
• Object diagram
• operation - the signature of a Method of a Class; consists of Visibility, Parameter list, Return Type and Property-string
• operator - an algorithmic feature of Interaction Frame that defines the behaviour of that frame. Examples include:
  • alt - multiple alternatives each with a Guard. Only one alternative can be true.
  • opt - single alternative with a guard condition.
  • par - each fragment is run in parallel.
  • loop - the fragment iterates according to a guard condition.
  • region - only one Thread can enter the fragment at a time.
  • neg - an invalid interaction.
  • ref - an interact defined in another diagram.
  • sd - inclusion of a complete Sequence diagram in the fragment.

• package -
• package diagram
• parameter - data passed in a Message to be used within the Method
• partition - section of an Activity diagram or Sequence diagram occupied by a single Class or Object
• pin - a parameter of an Action
• polymorphism - the ability of Objects belonging to different Classes to respond to Operations of the same name, each one according to the right Class-specific behaviour
• profile - Provides a generic extension mechanism for building UML models. Other extension mechanisms include stereotypes.
• property - an Attribute or an Association
• property-string - a qualifier for Attributes and Operations. Examples include {ordered}, {readonly}, {unique}.

• qualified association - tightens the multiplicity or role of an association between 2 classes with respect to a third qualifying class

• realization -
• reference -
• relation -
• return - a reply that may be issued from a Method following a Message
• role - description of the part played in an Association by one of the Classes in the Association

• scenario - a narrative describing foreseeable interactions
• self-call - a Message from an Object to one of its own Methods
• sequence diagram - describes the Messages sent between a number of participating Objects in a Scenario
• signal - an Event which can occur in an Activity diagram in three different ways: as a Time Signal, as a signal which an Activity can listen for and a signal which an Activity can send
• state - an Object exists at one of the States described in a State machine diagram
• state diagram - synonym for state machine diagram
• state machine diagram - describes the lifetime behaviour of a single Object in terms of in which it exists and the Transition between those States
• static attribute - an attribute that does not relate to a specific object but is at class level; that is, an attribute that is common among all objects of that class
• static modeling - use of class diagrams to describe structure
• static operation - an operation that does not relate to a specific object but is at class level
• Stereotype - a notation allowing the extension of UML symbols. Some are defined within Profiles. Common predefined UML stereotypes include:

• • Abstract - an abstract class
  • Actor
  • Access
  • Association
  • Become
  • Bind
  • Call
  • Copy
  • Create
  • Derive
  • Destroy
  • Document
  • Enumeration
  • Exception
  • Executable
  • Extend
  • Facade
  • File
  • Framework
  • Friend
  • Global
  • Import
  • Interface
  • Invariant

• • Library
  • Local
  • Metaclass
  • Model
  • Parameter
  • Postcondition
  • Powertype
  • Precondition
  • Process
  • Refine
  • Requirement
  • responsibility
  • Self
  • Send
  • Signal
  • Stereotype - Applies to a class. Shows that a class is actually a stereotype that can be applied to other elements
  • Stub
  • Table
  • Thread
  • Trace
  • Type
  • Use
  • Utility

• Structure diagram
• superstate - construct allowing several States which share common Transitions and Internal Activities
• Swim lane - synonym for Partition
• system model - The logical UML model being represented through one or more UML diagrams

• template - a Class that accepts a compile-time parameter defining the Type to be used within the Class; often to implement Collections of any Type
• Thread - a sequence of instructions whose execution is being scheduled by the Operating System and may run in parallel with other threads
• Timing Diagram
• token - symbolizes the Thread of an activity diagram
• transition - movement from one State to another in a State machine diagram. The transition is specified by its Trigger-signature /Activity
• type - the options are: an elementary Value type such as integer, string, date, or boolean or a Reference type defined in a Class

• use case - a technique for capturing functional requirements of systems
• use case diagram
• Uses dependency -

• visibility - the availability for access of elements in a model. Typically used to limit the visibility of features defined by a Class (e.g., attributes, operations). When appllied to features defined by a class, the standard options are:
  • private (-): available only within the Class in which it was defined. This is the most limited visibility
  • protected (#): available within the Class in which it was defined, and within any subclass of this class
  • package (~): available only within the Package which directly contains the defining Class
  • public (+): available to any Class that can see the defining Class. The is the least limited visibility.

• workflow -

• XMI - An OMG standard for exchanging metadata information via Extensible Markup Language (XML).

• Fowler, Martin ((Can we date this quote?)) UML Distilled: A Brief Guide to the Standard Object Modeling Language, 3rd ed. edition, Addison-Wesley, →ISBN

• Tom, Pender (2003) UML Bible, John Wiley & Sons, →ISBN