Speaking subskills are the cognitive operations a speaker carries out to turn a communicative intention into fluent, appropriate spoken language in real time. A competent speaker is doing several things at once: deciding what to say, finding the words and grammar to say it, articulating the sounds, and listening to their own output to catch and fix errors, all while managing the back-and-forth of a conversation with another person. These are not stages in a speaking lesson but parallel processes that run, mostly below awareness, every time someone opens their mouth.
The foundational distinction in the field comes from Martin Bygate's Speaking (Oxford University Press, 1987, in the Candlin–Widdowson teacher-education series). Knowing a language is not the same as being able to use it under the pressure of real-time production. A learner can know a grammar rule perfectly on a worksheet and still fail to deploy it in conversation, because spoken production demands that knowledge be retrieved and assembled fast enough to keep a conversation moving. Bygate frames speaking as a skill: knowledge that has been proceduralised to the point of being usable on the fly. This is why grammar accuracy on a written test predicts spoken accuracy so poorly, and why fluency has to be developed through speaking, not through learning more about the language.
Bygate divides oral skills into two families. Motor-perceptive skills are the basic capacities to perceive, recall, and articulate the sounds and structures of the language in the correct order. These are context-free: producing the right phoneme, stress pattern, or word order regardless of who is listening or why. Mastering them is necessary but nowhere near sufficient, because they treat speaking as the production of correct sound sequences rather than as communication.
Interaction skills are what turn motor-perceptive competence into communication. They cover the decisions a speaker makes about what to say, how to say it, and whether to develop or abandon a point, all in light of their intentions and their relationship with the listener. Bygate splits these into routines (conventional ways of structuring information, such as a narrative sequence or a set of directions) and negotiation skills (managing the interaction itself so that meaning gets through). The interaction skills are where most learners' real difficulty lies, and where traditional drill-based instruction does the least.
The cognitive machinery underneath these skills is best described by Levelt's (1989) model of speech production, which Bygate and later researchers applied to second-language speaking. Producing an utterance runs through four operations:
In a first language these run automatically and in parallel; in a second language they compete for limited attention, which is why pushing for complexity often costs accuracy or fluency and vice versa, the trade-off captured by the CAF framework. Judit Kormos extended Levelt's model to second-language speakers, modelling how slower lexical retrieval and first-language interference load the system.
The negotiation side of speaking has its own inventory of operations, drawn from conversation analysis:
When a speaker lacks the words for an idea, communication strategies such as circumlocution (describing around a missing word) keep the message alive; these are the practical face of strategic competence.
H. Douglas Brown's framework for assessment offers a complementary, more granular list. He splits oral production into micro-skills, the production of small units (phonemes, stress and intonation patterns, fluent runs of speech at varying rates, monitoring devices such as pauses and fillers), and macro-skills, the higher-order accomplishments (appropriate register and style, cohesion, conversational rules, accomplishing communicative functions, and using nonverbal cues). The micro/macro split parallels Bygate's motor-perceptive/interaction division: the small mechanical units underneath, the communicative accomplishments on top.
Christine Goh and Anne Burns, in Teaching Speaking: A Holistic Approach (Cambridge University Press, 2012), reframed the inventory for teaching. They argue that speaking is too often done in class but never taught, and that developing it means addressing three coordinated areas: language and discourse knowledge (vocabulary, grammar, phonology); core speaking skills (pronunciation, interaction management, discourse organisation); and communication strategies (the cognitive and metacognitive moves that help learners manage a speaking task). Their socio-cognitive stance treats speaking as both a mental process and a social act, and their teaching-speaking cycle scaffolds learners through planning, performing, and reflecting rather than simply throwing them into unsupported conversation.
Match the activity to the subskill. A task that pushes fluency (repeated retellings under decreasing time pressure) develops automaticity in formulation; a task that isolates interaction management (information-gap exchanges that force clarification) develops negotiation. Treating "speaking practice" as one undifferentiated thing leaves whole subskills untouched.
Build automaticity, do not just teach knowledge. Because the bottleneck is real-time retrieval, learners need volume and repetition under communicative pressure, not more rules. Task repetition is well evidenced as a way to free up attention for higher-quality output on a second attempt.
Teach interaction management explicitly. Turn-taking, negotiation, and conversational routines can be named, modelled, and practised. Learners who have the language but cannot manage the interaction stall in real conversations, and the fix is targeting the interaction subskill, not adding vocabulary.
Do not conflate subskills with lesson stages. Conceptualising, formulating, and monitoring run in parallel inside a single utterance; they are not the first, second, and third parts of a lesson. The teaching question is which process to relieve pressure on (planning time relieves conceptualising; pre-taught chunks relieve formulating) so the learner can devote attention elsewhere.