Code-Switching

Thinking Tools6.3 Implementing the Action Plan

From sociolinguistics: Jan-Petter Blom and John J. Gumperz gave code-switching its social meaning (1972), developed by Gumperz (1982) and Carol Myers-Scotton (1993).

Shifts vocabulary, register, depth, and references to suit each audience so the same substance is heard intact by engineers, executives, finance, or frontline staff.

One six-week slip, told four ways, with the same facts in each

The billing system's go-live moves from March 3 to April 14. At each of the lead's turns, pick the version for that group: the reply shows how it landed, and the verdict says whether it kept the facts and spoke the group's language. Illustrative.

Program lead

Audience

Program lead

Audience

Select a stage to read it, or play the sequence.To the engineers. Twelve engineers on the billing build. They know the tax engine and will want the failing tests, the patch date, and the retest plan.Pick a move to see the reply.Which tests? Which build? The engineers stop reading and message the tech lead for the facts.Executive language: nothing they trust, so they look for the substance elsewhere.The tech lead answers within the hour with a retest schedule for the week of February 24.Their code: tests, builds, and dates, and the same facts every other version carries.Engineers reply. The reply is about the work, not about the message. That is the sign the code fit.To the executives. The executive committee has an hour on Thursday. It needs the date, the cost, and whether anything needs its decision.Pick a move to see the reply.The chair asks what a rate test is, and twenty minutes go to the tax engine.Engineering language: true, but the committee hears detail and no answer.The chair notes the date and the cost, and the committee moves to the next item.Their code: the date, the money, the ask, with nothing softened.No questions. Six weeks later the committee learns the real size of it from finance.Their register, but the substance is gone: no date and no cost.Executives reply. A short reply with no follow-up questions means the three facts arrived whole.To finance. The controller needs to know what lands in which budget, and when.Pick a move to see the reply.The controller asks for the number in writing and copies the CFO.Softened: finance trusts amounts and dates, and this has neither.The controller books the accrual and asks for the vendor's invoice date.Their code: a line item, an amount, and a quarter.Finance replies. Finance acts on the number at once, which is what the message was for.To the billing clerks. Forty clerks will use the new system every day. They want to know what changes in their week.Pick a move to see the reply.Two clerks ask their supervisor whether their jobs are changing.Engineering language: true, and it tells them nothing about their week.The clerks ask which training slot they are in.Their code: the date and what it means for their work.Clerks reply. The question is about their next step, not the project. The message landed.

Versions that keep the facts and fit the group: 0 of 4Versions that keep the facts and fit the group: 1 of 4Versions that keep the facts and fit the group: 2 of 4Versions that keep the facts and fit the group: 3 of 4Versions that keep the facts and fit the group: 4 of 4

  1. To the engineers (Program lead). Twelve engineers on the billing build. They know the tax engine and will want the failing tests, the patch date, and the retest plan.
    • The program is moving to a revised timeline to protect quality at launch. Which tests? Which build? The engineers stop reading and message the tech lead for the facts. Executive language: nothing they trust, so they look for the substance elsewhere.
    • Cutover moves from March 3 to April 14: 4 of 30 rate tests fail; the patch lands Feb 24. The tech lead answers within the hour with a retest schedule for the week of February 24. Their code: tests, builds, and dates, and the same facts every other version carries.
  2. Engineers reply (Audience). The reply is about the work, not about the message. That is the sign the code fit.
  3. To the executives (Program lead). The executive committee has an hour on Thursday. It needs the date, the cost, and whether anything needs its decision.
    • The tax engine fails 4 of 30 rate tests, so cutover waits for the vendor's patch. The chair asks what a rate test is, and twenty minutes go to the tax engine. Engineering language: true, but the committee hears detail and no answer.
    • Go-live moves to April 14. A quarter more on the old system: $180,000. Nothing to decide. The chair notes the date and the cost, and the committee moves to the next item. Their code: the date, the money, the ask, with nothing softened.
    • We are tracking a small adjustment to the go-live schedule. No questions. Six weeks later the committee learns the real size of it from finance. Their register, but the substance is gone: no date and no cost.
  4. Executives reply (Audience). A short reply with no follow-up questions means the three facts arrived whole.
  5. To finance (Program lead). The controller needs to know what lands in which budget, and when.
    • The delay should be cost-neutral overall once the new system's savings start. The controller asks for the number in writing and copies the CFO. Softened: finance trusts amounts and dates, and this has neither.
    • Old-system license runs one more quarter: $180,000, unbudgeted, booked in Q2. The controller books the accrual and asks for the vendor's invoice date. Their code: a line item, an amount, and a quarter.
  6. Finance replies (Audience). Finance acts on the number at once, which is what the message was for.
  7. To the billing clerks (Program lead). Forty clerks will use the new system every day. They want to know what changes in their week.
    • Cutover is deferred pending tax-engine regression results and the vendor's patch. Two clerks ask their supervisor whether their jobs are changing. Engineering language: true, and it tells them nothing about their week.
    • New system starts April 14, not March 3. March is unchanged; training moves to April 7. The clerks ask which training slot they are in. Their code: the date and what it means for their work.
  8. Clerks reply (Audience). The question is about their next step, not the project. The message landed.

An example

A program lead at a regional insurer had one piece of news for four groups on the same Thursday: the new billing system would go live on April 14, six weeks after the March 3 date everyone had planned around, because the tax engine still failed four of thirty rate tests. Her first draft was one email to everyone, written like her weekly status report to the engineering leads.

She wrote four versions instead. The engineers got the failing tests, the vendor's patch date, and the retest plan. The executive committee got three sentences: the new date, the cost of running the old system for one more quarter, about $180,000, and no decision needed. Finance got the unbudgeted license line and the quarter it would land in. The billing clerks got what mattered to their week: March unchanged, training moved to April 7. Before sending, she checked that every version carried the date, the cause, and the cost.

Why it works

Every working group has a dialect: the words it uses for itself, how formal it expects a message to be, how much detail it wants, and which kind of evidence it trusts. A message in someone else's dialect makes the listener translate, and most people stop translating early. They hear an outsider and discount what follows. Switching into the listener's code removes that cost, so their attention goes to the substance.

The discipline that makes it more than tact is the substance check. Every version carries the same facts; only the delivery changes. Without that check, adapting to an audience slides into telling each group what it wants to hear, and the groups compare notes. With it, four messages say one thing, and each lands the first time.

The thinking behind it

Jan-Petter Blom Jan-Petter Blom, a social anthropologist at the University of Bergen, studied with John J. Gumperz how people in a northern Norwegian community moved between their local dialect and standard Norwegian. Their paper, Social Meaning in Linguistic Structure: Code-Switching in Norway (1972), showed that the choice of code carried social meaning of its own: which dialect a speaker used told listeners what kind of exchange they were in. It is one of the founding works on code-switching.

John J. Gumperz John J. Gumperz, a Berkeley sociolinguist, founded interactional sociolinguistics. With Blom he separated situational switching, where a change of setting brings a change of code, from metaphorical switching, where a speaker changes code to change the meaning of the moment. In Discourse Strategies (1982) he showed how small contextualization cues in speech tell a listener how a remark is meant to be taken. Tuning a business message to each audience is a later, practical use of the idea that no one person authored.

Carol Myers-Scotton Carol Myers-Scotton, an American linguist, set out the markedness model in Social Motivations for Codeswitching: Evidence from Africa (1993). Speakers choose a code to signal the rights and obligations they claim in an exchange. The expected code is the unmarked choice; switching away from it is marked, and it says something about the relationship. In her terms, a lead who answers a clerk in engineering jargon is making a marked choice, whether she means to or not.

How to run it

  1. Fix the substance first. Write down the three or four facts every version must carry, such as the date, the cause, and the cost, before drafting any version.
  2. Profile each audience. Note the words they use for themselves, how formal they expect it, how much detail they want, and the evidence they trust.
  3. Mark where your default departs. Read your first draft as each group would and underline the terms, length, and evidence that belong to your own team.
  4. Rewrite in their code. Match their vocabulary and level of detail, draw examples from their work, and lead with what changes for them.
  5. Run the substance check. Lay the versions side by side and confirm each carries every fact, with nothing softened or dropped.
  6. Test with a native speaker. Ask one member of the audience whether they could have sent the message themselves without wincing.
  7. Watch the reply. If people question the delivery more than the content, the code is still off; adjust the next message.

Where it goes wrong

The first failure is partial fluency: the audience's buzzwords pasted onto the writer's own structure and detail. It reads as imitation, which costs more trust than plain speech would. Ask someone from the group to read it before it goes.

The second is drift in the substance. The executive version says "a small schedule adjustment" while the engineers hear six weeks, and the gap surfaces when the groups compare notes. Keep the fact list beside the drafts and check every version against it.

The third is switching so far that the writer stops sounding like herself, which reads as a performance. Match the listener's level and evidence, and keep one's own voice.

When to use something else

  • Persona Shifting. When the message is settled and the open question is how the decision looks from each stakeholder's seat.
  • Pyramid Principle. When the trouble is structure for every audience at once: the answer first, then the grouped support.
  • Culture Map. When the gap is national culture, such as how directly people disagree or how they build trust.

Where it sits in SOLVE eX

Go to the source

  • Social Meaning in Linguistic Structure: Code-Switching in Norway, Directions in Sociolinguistics, 1972
  • Discourse Strategies, 1982
  • Social Motivations for Codeswitching, 1993

Back to the catalog