Intellectual Practices of Keith Lyons
Clyde Street is full of new terms, metaphors, labels and ways of describing things.
The whole-corpus study retained only twenty-six records as positive evidence for this particular operation.
That gap became the puzzle.
What had to happen before a change in language counted as a repair rather than simply another phrase, metaphor or way of talking?
A preposition can carry a different relationship
On 25 October 2014, Keith Lyons wrote a short post called Moving from for to with.
He had spent the day talking with people about performance environments in sport. Those conversations returned him to what he described as a recurring question across thirty-five years of performance analysis: when should analysis move from being done for the coach to being done with the coach?
Keith did not merely add the word with to his vocabulary.
He contrasted two relationships.
Analysis for the coach leaves the analyst primarily in a follower role. Analysis with the coach opens the possibility of analysts leading conversations at important moments and, eventually, players leading conversations about performance themselves.
Keith then made the shift explicit. He wrote that he wanted to celebrate “the possibilities of with”.
For the study, this was a strong case because the change was local enough to see.
A specific organising relation had been identified. Professional experience and recent conversations had made the old relation insufficient for what Keith wanted to describe. He replaced it with another relation and explained what that replacement changed.
John Dickson’s writing on humility appears in the post, but the ownership line remains clear. Dickson owns his account of humility. Keith owned the for-to-with reformulation of his own professional practice.
This is the kind of case that kept the finding narrow.
The issue was not whether Keith’s thinking had changed in some broad sense. It was whether a term, distinction, notation or representation had been exposed as a poor fit and Keith had visibly repaired it.
The next post shows why that distinction mattered.
The same words can belong to a different operation
One day later, Keith published Extending the discussion of With.
The post takes the for-to-with distinction and develops a three-step account: solitary heroic work, independent peers facing similar challenges, and a shared condition of being “in it together”. A cycling incident becomes the central representation for shared responsibility and trust.
The language of for and with is still present.
But the whole-corpus study did not treat this second post as another instance of the same narrow language-repair operation. It retained it as a broader change in the inquiry.
That boundary is important.
Once the problem expands into a new conceptual structure, the relevant question is no longer simply whether one term has been repaired. The inquiry itself is being reorganised.
So even adjacent posts using the same vocabulary can belong to different findings.
Language repair was not a synonym for conceptual development.
A familiar label can fail because the scale is unknown
A second kind of repair appeared later in the same year.
Keith was preparing the open #UCSIA15 course. The familiar abbreviation MOOC carried the word Massive.
But Keith did not know how many people would register.
He therefore reworked the familiar label in his own account. The M would stand for Modest.
The interesting part is not the joke.
The source gives the change a condition. The course scale was unknown, and Keith wanted the label to reflect that uncertainty and the character of the course he was building.
That is why the study distinguished this from ordinary wordplay.
The representation was being adjusted to fit what was actually known.
Uncertainty itself did not make this a language-repair case. A limitation can constrain a claim without changing its terminology. Here the consequence was different: the uncertainty prompted Keith to alter the organising label.
Repair can mean redefining rather than abandoning a word
Not every qualifying case replaced one word with another.
In June 2014, Keith encountered Stephen Downes’s rejection of the term curation.
Downes argued that curation suggested selection, filtration, passivity, preservation and objectivity, whereas his own work involved active engagement with authors and ideas.
Keith did not simply adopt Downes’s rejection.
He described the critique as leaving him with a quandary and then re-specified what curation meant in his own practice. His curation, he wrote, involved engagement and active choice. He described it as fallible, partial, serendipitous, subjective, reflective and dynamic.
That made the case useful for the whole-corpus study.
The repair did not require Keith to discard the contested word. He could keep it while changing the account of what the word was allowed to mean in relation to his practice.
Again, ownership matters.
Sue Waters owns her account of curation. Stephen Downes owns his critique. Keith owned the decision to retain and redefine the term in his own description of what he was doing.
A borrowed objection is not itself evidence of Keith’s language repair.
The repair appears when the record shows Keith doing something with the mismatch in his own account.
A role name can become too small for the work
By April 2017, another kind of mismatch had become explicit.
Keith had been travelling around England visiting coaches in a critical-friend group. The journeys led him to reflect on personal learning and on how he described his own role.
He wrote that, if he placed so much emphasis on coaches’ personal learning, he could “move on from the descriptors ‘coach developer’ and ‘mentor’.”
He then settled on another descriptor: “travel companion”.
This could easily have been dismissed as metaphor or rebranding.
The study retained it because Keith explicitly identified the inherited descriptors as no longer fitting the work he was trying to describe. He then specified what the replacement was meant to hold: leading, following, guiding, meddling and accompanying an entangled personal-learning journey.
The metaphor earns its place because a mismatch has already been made visible and the replacement is doing representational work.
That is different from simply finding an attractive image.
The distinction matters because Clyde Street contains a great deal of metaphor.
Most metaphor did not qualify.
Policy aspiration can be a language problem too
In January 2018, Keith was thinking about sport systems and the language of aspiration.
Two pieces by Thomas Oppong and David Weinberger brought him back to optimisation. Keith placed that idea against the rhetoric of being “world best” and “world leading”.
He wrote about sport systems flourishing without the burden of those outcome labels and developed optimisation through ideas such as good enough, best effort, learning, playfulness and transparent fairness.
The external ideas remain externally owned. Oppong, Weinberger and the authors they cite supply arguments about progress, optimisation and imperfection.
The Keith-owned move is the repair of his own representation of what a sport system should aspire to be.
The study retained the case because the old aspiration language is explicitly placed under pressure and an alternative representational frame is developed in its place.
That does not establish that optimisation is a universally better word.
It shows that, in this source, Keith found the existing rhetoric inadequate for the conditions he wanted to describe.
A representation can be repaired by adding something it now needs
The latest approved positive came on 12 September 2019.
Keith was thinking about the verbs used to describe work in performance analysis and data.
He looked back to the pre-digital representation he had worked with:
Observe.
Record.
Analyse.
Model.
The role, he wrote, had changed in a digital era. He considered IBM’s AI Ladder and Peter Dowrick’s work on feedforward and mental simulation.
Then he altered his own representation.
“I would add to this feedforward.”
This case broadened the form of the finding without broadening its boundary.
The repair did not consist of replacing one adjective or noun. It changed an organising set of professional verbs so that the representation could include prospective, future-oriented work under changed digital conditions.
IBM owns its AI Ladder. Peter Dowrick owns his feedforward research. Keith owned the revision of the professional representation he had been using for his own field.
The study did not infer that artificial intelligence caused a general change in Keith’s thinking. It retained a much narrower claim: this source documents a specific representational revision under changed digital and AI-related conditions.
Talking about language is not enough
The negative cases were essential because Clyde Street often uses language itself as an object of inquiry.
In November 2010, for example, Keith wrote Aural Triggers for a Language of Performance?
He drew on music, sound recording and orchestral performance to think about observation, listening, mentoring and sporting performance.
This is rich material about language.
But the whole-corpus study did not treat it as a strong language-repair case.
The source connects music and sound to sporting performance and lets those distant domains do analytical work. What it does not clearly show is a specific Keith-owned term, notation or representation being identified as inadequate and then visibly repaired.
That distinction prevented the finding from swallowing every post that explored vocabulary, analogy or metaphor.
A language of performance is not automatically language repair.
The early cases show why calibration mattered
The earliest approved positive, from October 2008, also stayed deliberately weak.
While reading about learning design, Keith noticed a reference to trial-and-error learning and asked whether affordance might be better supported by “trial and improvement” and dynamic evaluation.
The move is visible, but it is tentative.
Keith posed it as a conjecture. The post does not establish a developed before-and-after terminology system, and several other intellectual operations were happening around it.
The study therefore retained the case only at Possible strength.
That caution matters to the whole-corpus finding.
All nine Possible positives occur in the early 2008–2010 decision layer. Later approved cases are Explicit or Probable. But the study explicitly refused to turn that distribution into a story of increasing sophistication or commitment.
The early coding architecture was looser. Calibration standards changed. Later cases were often reconstructed with more source detail and stronger QA. The apparent shift in evidence strength is therefore method-sensitive.
The safest statement is not that Keith became better at repairing language.
It is that later cases often make the mismatch and repair easier to locate.
What the whole corpus established
Across the frozen 2,051-record Clyde Street corpus, twenty-six records were retained as positive evidence for this operation.
Nineteen were single-primary decisions. Seven were early multi-code records that contained the operation.
Thirteen positives were Explicit, four Probable and nine Possible. The earliest approved positive is from 21 October 2008. The latest is from 12 September 2019. Approved positives occur across eighteen chronological batches.
This is a recurrent finding, but a sparse one.
The distribution contains substantial chronological gaps. No approved positive occurs after “Describing What We Do” (12 September 2019). The study treats that as descriptive silence, not evidence that the practice disappeared.
The forms vary: trial-and-improvement language, course labels, notation, curation, relational prepositions, professional-role descriptors, coach-learning language, policy aspiration rhetoric and representations of analytical work.
No one mismatch type explains them all.
A draw can expose a category failure. A colleague can challenge a term. Uncertain scale can make a label overclaim. Professional experience can make a role description feel wrong. A historical subject can correct wording. Digital conditions can make an inherited representation incomplete.
What joins the cases is not the topic.
It is the local operation: a specific representation no longer fitted, and Keith visibly changed it.
What did not count
The boundary portfolio contains nineteen explicitly routed rivals and near-misses. It is not a denominator of every Clyde Street record that failed to qualify. It is a deliberately retained set of cases showing how the finding could have been inflated.
The recurrent failure modes are revealing.
Broad conceptual change did not qualify simply because new language appeared inside it.
Metaphor did not qualify merely because it was vivid.
A new term did not qualify if no mismatch had first been exposed.
A proposed future wording change did not automatically count as an enacted repair.
An external definition or terminology system remained externally owned unless Keith performed a distinct repair in his own account.
Repeated use of a label after it had already been changed did not generate another positive simply through serial reuse.
And practical action belonged elsewhere when the main consequence was what Keith did rather than how he repaired the representation.
These controls are why the finding remained narrow even though Clyde Street is linguistically inventive.
What the finding permits
The whole-corpus evidence supports a high-confidence claim that narrow mismatch-and-repair recurs across multiple periods and domains in Clyde Street.
It also supports a high-confidence boundary around the later Explicit cases and around ownership.
Keith did not have to invent the pressure that exposed the mismatch. Another person, a policy, an event, a dataset or a technological condition could do that.
But the repair counted as Keith-owned only when Keith visibly changed his own terminology, notation or representation.
The finding also permits a more modest developmental description: the later corpus often contains repairs whose before-and-after language is directly locatable, whereas many early cases remain tentative naming or distinction-making.
It does not permit that difference to become a claim of chronological intensification.
What the finding does not permit
The evidence does not show that Keith habitually renamed everything.
It does not show that new language is inherently better than inherited language.
It does not turn every metaphor, reframing or vocabulary innovation into a finding.
It does not establish that every repaired term changed practice, improved decisions or affected other people.
It does not establish that the people whose criticism or ideas exposed a mismatch caused the eventual repair unless the source shows that relation.
It does not justify treating language repair as a personality trait, a universal method or one stage in a larger fixed process of inquiry.
And it does not permit the late digital and AI-related case to be read backwards as though artificial intelligence explained the earlier chronology.
The bounded finding is simpler.
Across Clyde Street, there were recurrent occasions when Keith did more than change his mind or find a better-sounding phrase. A particular representation was shown to fit badly, and he repaired the representation itself.
Why the distinction matters now
The easy lesson would be to encourage constant renaming.
The evidence points in almost the opposite direction.
New words are cheap.
A genuine repair needs a reason.
Something has to expose the mismatch: an event the categories cannot hold, a relationship the old preposition misrepresents, a scale the inherited label exaggerates, a professional role that no longer describes the work, or a changed technical condition that leaves an older representation incomplete.
Even then, the new wording is not proven correct forever.
It is a revised representation under particular conditions.
That is why the whole-corpus boundary remains useful. It asks us to separate verbal novelty from evidentially motivated repair.
The question the finding leaves with the reader is therefore:
Question to carry
What evidence would distinguish a genuine repair in my language from a new phrase that simply feels better?
Sources followed
Keith Lyons, “Moving from for to with” — 14/10/25
Keith Lyons, “Extending the discussion of With” — 14/10/26
Keith Lyons, “#UCSIA15 — A Modest Open Online Course (MOOC)” — 14/12/09
Keith Lyons, “Curation and Kindness” — 14/06/18
Keith Lyons, “Travel Companion 1” — 17/04/27
Keith Lyons, “Optimistic About Optimisation” — 18/01/30
Keith Lyons, “Describing What We Do” — 19/09/12
Keith Lyons, “Aural Triggers for a Language of Performance?” — 10/11/28
Keith Lyons, “CCK08 Week 7: Affordance? Trialing, Improving/Mashing” — 08/10/21