How do you translate “did so” accurately when English leaves an earlier action unspoken? Begin with a simple sentence: “The researcher checked the figures, and the editor did so too.” The editor checked the figures. The words did so do not introduce an unrelated task; they refer to an earlier action. This pattern is known as do-so anaphora, a form of verbal or event anaphora. For translation, the goal is to keep the action, participants, time and any relevant limits intact, even if the target language has no equivalent of English did so.
Readers searching for do so meaning, how to translate did so, do so versus do it, verb phrase anaphora examples, or how to translate actions without repeating the verb need an answer that works outside grammar exercises. A fluent translation can be wrong if it turns “the editor did so too” into “the editor approved the figures,” “the editor ordered it,” or “the editor did the same work at the same time.” None follows automatically from the source. To translate reliably, recover the smallest action that the discourse supports, identify its participants and decide what is genuinely shared.
This guide explains do so, does so, did so, doing so and to do so with everyday examples, professional documents and source-to-target quality checks. It distinguishes event anaphora from ordinary auxiliary ellipsis, nominal reference and null complements; shows how modifiers and comparisons can change the scope of the repeated action; and teaches a practical method for human or AI-assisted translation. The key principle is to repeat meaning when necessary, without inventing a new event.
A one-minute method for resolving did so
Read the preceding clause first. Identify the main action in ordinary language. In “The volunteer scanned the labels, and the archivist did so afterward,” the recoverable action is scanning the labels. Next identify the subject of the do-so clause: the archivist. Then identify any new information that accompanies do so: afterward says the second event happened later. A safe expanded paraphrase is “The archivist scanned the labels afterward.” The original does not specify which scanning equipment was used or whether the scans passed quality assurance.
For translation, you can use the target language’s natural way to indicate a repeated action, or you can repeat the verb. There is no requirement to preserve the English words do and so as independent pieces. In some languages, a repeated verb is clear and idiomatic. In others, a verbal proform, omission or short reference expression may carry the same event. Choose the construction that a reader can interpret without guessing.
The easy quality-control test is to write two propositions. First participant performs P under the first clause’s stated conditions. Second participant performs P under the second clause’s stated conditions. Then ask which conditions truly carry over. Do not assume that every surrounding modifier is repeated: a following contrast such as “but in a different room” may explicitly override location. Resolve these choices in context, not by mechanically copying all words from the antecedent.
A useful working question is “What exactly did the second person do, and which details do we know about that second action?” If the answer requires an invented location, motive or outcome, the translation has exceeded the source.
What do so means, and why it is not merely a synonym for do it
Cambridge English grammar describes do so as a substitute that typically refers to an action already mentioned, often in a somewhat more formal register than do it or do that. This explains why instructions, reports and formal notices frequently use “to do so,” while conversational speakers may simply repeat the verb. But stylistic differences are only part of the story. Verbal reference with do so has been studied as an event-related construction with its own semantic and discourse behavior.
Compare “Mira opened the packet, and Jo did so too” with “Mira opened the packet, and Jo did it too.” Both may refer to opening the packet in an ordinary setting. Yet do it may more readily behave like a reference to a task or particular act in contexts where do so sounds formal, and acceptability can differ with the kind of predicate being referred to. A translator should not simply construct a fixed bilingual lookup chart that treats all forms as perfectly interchangeable.
Likewise, “do that” can include a demonstrative contrast or point to a particular action previously described. “Do the same” may invite comparison at a different level of detail: same task type, same method or same object, depending on context. These expressions are close relatives, not always exact copies of one another. Their differences matter when the source signals contrast or when the target language uses different reference strategies.
For reference, see the Cambridge Dictionary grammar guide to do and Michael Houser’s research on do-so anaphora. The practical lesson is that translation should follow the event being picked out, rather than equating every two-word substitute with the same target phrase.
Recovering an event rather than simply copying a visible verb
An antecedent does not always present a neat verb phrase that can be copied without interpretation. Imagine: “After months of planning, the team finally secured permission to access the archive. The partner organization did so a week later.” The obvious repeated event is securing permission, not planning for months. The phrase did so points readers toward an event that is prominent in the discourse. The modifier a week later refers to the partner organization’s success, not to a repetition of the original month-long planning period.
This is a strong reason to analyze *event reference* and not merely the nearest verb on the screen. A rushed translator may choose access rather than secure permission, because both are nearby. But having permission to access an archive is not the same as actually accessing it. The target needs to retain that distinction, especially in records of rights, licensing or institutional approval.
A second example: “The pilot project brought three groups together to agree on the revised procedure, and the later workshop did so on a smaller scale.” What did the workshop do? Context may license a broad event of bringing groups together for agreement, rather than one isolated lexical verb. An overly narrow translation such as “the workshop revised the procedure” could report an action that the source does not independently establish.
The safest approach is to identify the relevant discourse event and draft a plain-language expansion. Check whether it requires the same outcome, participants, method or objective. If the source could still be true without one of those details, avoid hardwiring that detail into the target. A verbal anaphor can point to a complex activity, but not every aspect of the earlier scene necessarily transfers.
Who did what: participant roles before stylistic polish
“The technician cleaned the filter, and the supervisor did so afterward.” Here technician and supervisor are two different agents; cleaning the filter is the shared action type. But if there are several filters, the sentence’s context determines whether the supervisor cleaned the same filter or another relevant filter. Do not add “the identical physical filter” without a reason. Nor should the source be reduced to “the supervisor checked the filter.” Cleaning and checking are different.
In a more complex sentence—“The team helped the residents complete the form, and the volunteers did so during the evening session”—the repeated action likely includes helping residents complete the form. The volunteers are the second helpers. The residents remain beneficiaries or recipients of help, not automatically the second agents. Translations that change who helps whom create real operational misunderstandings.
Try writing a role grid: first agent, first action, first object or recipient; second agent, repeated action, object or recipient if established; new modifier. For the form example, team = first helper, volunteers = second helper, residents = people being helped, form = task object, evening = second event’s time. You can now put the words in target-language order without shifting roles.
The grid is especially valuable in languages where subjects are often omitted or where pronouns agree differently from English. The final target may be concise; your working analysis should be explicit. If a bilingual colleague reads the target and cannot tell who carried out the repeated action, give back some of the compression instead of trusting context to solve everything.
Time, aspect and modality survive the substitution
“Sara filed the appeal, and Omar did so yesterday” reports two filing events with at least one specified temporal detail. “Sara will file the appeal, and Omar will do so next week” reports expected future actions. “Sara could file the appeal, and Omar could do so too” reports ability or possibility, not necessarily filing. A translation that uses the same completed-past expression for all three versions erases their practical differences.
An auxiliary or modal around do so contributes information beyond action identity. May can mark possibility or permission according to context; must may impose obligation or support an inference; should can indicate advice, expectation or normative force. Being able to reconstruct file the appeal does not excuse ignoring these explicit meanings. The repeated predicate and the modal environment must both be translated.
Aspect also matters. “They have done so” presents the event in a perfect construction tied to the speaker’s temporal viewpoint. “They were doing so” presents an ongoing event or activity. “Having done so, they submitted the results” makes the completed previous action relevant to what follows. In a target language without matching auxiliary forms, use the appropriate local tense, aspect or adverbial structure. That is a translation decision, not word-for-word substitution.
To audit, imagine a world where the action has not yet occurred. A sentence with could or may can still be true, while a sentence asserting did may not. Imagine a world where the action has begun but not finished: an ongoing form may fit, while a completed form might not. These quick counterexamples reveal when a target has quietly changed an obligation, possibility or outcome.
Do so and stative verbs: a useful tendency, not an absolute slogan
In familiar English, do so often fits eventive actions better than stable states. “The athlete ran five laps, and the coach did so too” sounds easier than a constructed sentence such as “The athlete knew the answer, and the coach did so too.” English more naturally uses “the coach did too” for knowing. But the semantic restriction on do so has been debated, and corpus evidence shows that acceptable uses depend on construction, context and other factors.
Michael Houser’s dissertation examines agentivity, event type and stativity, and reports that certain syntactic environments can improve the acceptability of uses that otherwise sound odd. This is a caution against simplistic teaching rules such as “do so can never refer to a state.” The everyday tendency is real and useful, but language data are not always captured by one categorical test.
For translators, a bigger practical point is to ask what kind of meaning the source actually expresses. If the source is “she knew the answer, and he did too,” the target must preserve knowledge, not convert it into an action of learning or discovering merely because a target form that resembles do so prefers active events. A change from know to learn moves from state to change of state. The target can choose a different anaphoric construction.
The source may itself be nonstandard, rhetorical or deliberately unusual. In a novel, preserving a speaker’s odd register can matter. In a technical guide, a natural target that repeats the state predicate is usually best. Linguistic analysis helps explain the choices; it does not authorize a translator to change the stated relation just to make one proform fit.
Worked example: reviewing invoices without approving them
A procurement record says, “The accounting team reviewed the invoices on Monday. The external auditor did so on Thursday.” The repeated event is reviewing invoices. The second participant is an external auditor; its time is Thursday. The wording does not say the auditor approved the invoices, paid them, reconciled every amount, used the same checklist or reviewed every invoice in the entire organization.
Now suppose a translator writes, “The external auditor approved the invoices on Thursday.” That sounds plausible in procurement vocabulary, but the event has changed. A review may end in approval, rejection, recommendation or a request for more evidence. The source only establishes the reviewing action. Another mistranslation would say that the auditor checked them on Monday, copying the first event’s day and erasing Thursday.
A careful target can say, “The external auditor also reviewed the invoices, on Thursday.” If the local language allows a concise action anaphor, use it only if it preserves reviewing as the action. Notice also that “the invoices” may refer to a defined group in the prior sentence. If the second sentence says “a different group of invoices,” the translator must update that referent rather than blindly carrying the first object across.
A revision team can test the passage with two questions: What did the external auditor do? When did they do it? Correct answers are review the invoices and Thursday. A third question—did the auditor approve payment?—should remain unanswered. If the target provides yes, the translation has invented an outcome.
Worked example: translating safety instructions
Consider a notice: “Operators must isolate the power supply before cleaning the machine. Contractors working on the same equipment must do so as well.” Here do so refers to isolating the power supply before cleaning the machine, including the temporal condition. It is not enough for the target to say contractors should isolate power at some unspecified time; before cleaning is part of the safety-critical instruction.
A short sentence can carry a large dependency. The translator should break the source into action = isolate power supply; timing = before machine cleaning; affected role = operators and contractors; obligation = must. “Contractors must clean the machine” would recover the wrong verb. “Contractors may isolate the power” would weaken the requirement. “Contractors must turn the machine off at the end” would change the required procedure.
If the target audience needs an explicit repetition rather than a vague proform, write the full instruction, subject to the organization’s approved terminology and safety review. Avoid adding your own lockout procedure, voltage specifications or test steps absent from the source. A translator’s job is to preserve the existing instructions accurately, not to improvise safety guidance.
In safety-critical material, professional review is indispensable. The grammar check is the starting point: does the second sentence require the same action at the same relevant stage? Do not assume that a fluent rendering with a word meaning do guarantees that readers will understand the necessary timing and scope.
Worked example: meeting decisions and administrative authority
An institutional memo reads, “The committee considered postponing publication. The board did so at its next meeting.” At first glance do so seems to refer to considered postponing, not postponed publication. This distinction matters: considering an action does not entail approving or implementing it. A target that says “the board postponed publication” may misreport a decision.
But the preceding context can complicate recovery. If the text previously says that the committee voted to postpone, the most prominent prior event might be the vote or the postponement, depending on what the next sentence emphasizes. A mechanical nearest-verb rule cannot guarantee the correct analysis. Read the full paragraph and identify what kind of event “did so” is describing.
A source that intends exact formal decision status should ideally make the act explicit. Translators cannot rewrite source minutes as if they were free to settle ambiguity. If the distinction could affect governance, flag it for author clarification. For ordinary descriptive prose, a safe expansion is possible when context clearly identifies whether it is consideration, voting or implementation.
Use a decision-state grid: discussed, proposed, supported, approved, implemented. These are distinct. The phrase did so can refer to one of those states if the earlier discourse establishes it. Your target must match that state, not the next one the process might logically lead to. A readable translation is one that keeps the record auditable, not one that seems impressively decisive.
Worked example: research methods and replication claims
A research article reports, “The first laboratory measured samples at three time points. The second laboratory did so using a different instrument.” The recoverable method includes measuring samples at three time points, but the modifier using a different instrument explicitly distinguishes equipment. A translator must not assert both laboratories used the same instrument or infer that their measurements agreed.
Depending on how closely the study identifies its samples, the source may mean the second laboratory measured the same samples or corresponding samples of the same type. That reference needs contextual verification. Do so guarantees a connection to a prior activity, but it does not by itself certify physical identity of all objects or experimental equivalence of results.
A precise translation might repeat “measured samples at three time points” and then add the different-instrument clause. If word limits demand a shorter sentence, use the target language’s anaphoric form only if readers still recover the three time points. The timing may be central to the experimental design. Omitting it could turn a longitudinal method into a single measurement.
For verification, write a methods table with columns for laboratory, measurement type, sampling schedule, instrument and observed outcome. Fill only what the source states. The second laboratory can share the first three fields while differing in instrument; observed outcomes remain unknown. This separates action anaphora from methodological assumptions and protects research integrity.
Worked example: elections and public statements without invented results
A historical note says, “The candidate pledged to release the financial records. Her successor later did so.” In this context, do so can refer to releasing the records rather than pledging to release them, but the reading depends on the discourse and the chosen antecedent. If the source means the successor actually released the records, the translator should preserve that completed act. If the note explicitly says “made the same pledge,” then the repeated event is pledging, not releasing.
This pair is a useful caution because a preceding clause contains two verbal events: pledge and release. One is a communicative commitment, the other is the action promised. A smooth AI-generated sentence may leap to the most newsworthy one. The translator should instead inspect the source grammar, context and subsequent evidence.
In a different passage, “The candidate announced a review, and her successor did so too” most naturally reports an announcement, not necessarily a completed review. Any target that reports publication of a final review would need additional evidence. Likewise, naming a government office, election outcome or release date requires source support; the exercise cannot supply real-world facts.
The ethical test is whether a reader could cite the translated sentence as proof that a promised action occurred. If the source only reports a promise or announcement, the target must not produce such proof. Do-so recovery becomes an evidence problem when the antecedent includes layered actions.
Worked example: customer service escalations
A support record states, “The first-line agent escalated the ticket after checking the logs. The night-shift team did so when the alert returned.” Which action is repeated? The context may favor escalated the ticket after checking the logs as a bundled activity, or a narrower escalation event. A translator should examine whether when the alert returned replaces or supplements the earlier timing and whether the night shift checked logs as well.
A careless target could say “the night-shift team checked the logs” and omit escalation. Another could assert that both teams used identical workflows even though the sentence only refers to a repeated operational response. In a helpdesk system, checked, escalated, assigned, resolved and closed are distinct ticket states.
Where multiple antecedent actions compete, annotate the ambiguity and check related timestamps or case notes if available. Do not turn a probable reading into an unsupported certainty. If the writer can be consulted, ask which action did so is intended to denote. A short clarification can prevent a ticket history from becoming inconsistent across languages.
The source-to-target invariant is not the grammatical brevity of did so. It is the case event and the state transition associated with it. A reader in the target language should be able to distinguish escalation from analysis, and both from resolution, just as a reader of the complete source record can.
Worked example: accessibility settings and user instructions
A user guide reads, “You can enlarge the interface text in Settings. To do so, open Display and choose Text Size.” Here to do so points forward in the instruction sequence as a reference to enlarging text. It is not a generic permission to change every display property. Translating it as “to change the display” broadens the goal; translating it as “to make the screen brighter” changes the feature.
Not every instruction can be shortened safely. In a target language with long distance between the goal statement and the following steps, repeating “to enlarge interface text” may be more usable than a compact equivalent of to do so. Technical translation should optimize for reader understanding and correct task execution, not word-for-word fidelity to a short English form.
If a product uses controlled menu labels, retain their approved localized wording. Display and Text Size may have official translations chosen by the software team. But do not mistake those labels for the substantive event. The event is increasing text size; the labels identify how the user reaches the setting.
The same method works for privacy controls, language switching and notifications. Identify the action, verify the target UI labels, and test whether a reader can execute the stated task without accidentally changing another setting. If two preceding actions are listed, a later to do so may need explicit expansion to resolve ambiguity.
Worked example: education and different learning activities
A parent report reads, “Anya revised the vocabulary list twice before the quiz, and Joel did so during the weekend.” The natural repeated action is revising the vocabulary list, with a possible question about whether twice is included in what Joel did. The weekend modifier specifies when Joel revised. The translator should not automatically say Joel took the quiz, learned every word or scored the same result.
Notice how an event modifier can be part of the action description or part of the first participant’s circumstances. The phrase twice before the quiz describes Anya’s revision schedule. In some discourse settings the second event inherits the repetition count; in others the author merely reports that Joel revised too, with his own timing specified. If the count matters, source clarification may be needed. Do not silently insert “twice” unless the second sentence clearly supports it.
A learning record may have separate evidence for exposure, practice, retrieval accuracy and assessment performance. “Did so” is not a shortcut for mastery. Repeated practice does not guarantee retention, and an account of revision should not turn into a score claim.
To check the translation, ask what each child actually did, which schedule details are explicit for each, and what the source leaves unknown about learning. A reader should understand the repeated study activity without inferring that both students achieved identical results.
Worked example: travel documentation and one procedure
Consider the itinerary note, “The tour leader registered the group at the entrance, and the assistant did so at the second site.” The repeated procedure is registering the group, performed at another location. A target that writes “the assistant entered the second site” changes registration into physical entry. A target that writes “the assistant registered at the first site” loses the stated contrast.
The verb register can refer to signing in, submitting details, collecting entry passes or recording attendance depending on local practice. Translate the kind of registration documented by the source, not a more specific procedure invented from expectations about the venue. If the source provides an official system or form name, reproduce its verified localized label.
Now consider whether group refers to the same travelers at both sites. In a continuous trip narrative that is likely, but if the group has split, “the group” may need careful reading. Do so connects the two registration actions; it does not define the group’s membership in every situation.
An itinerary translator should track participants, venue, task, time and any new restrictions. That role-and-event table is often more useful than a dictionary entry for so. It makes short anaphoric instructions resilient when the target language uses different word order or omits pronouns.
Worked example: legal and compliance language
An internal report says, “The organization recorded the complaint and later notified the regulator. Its subsidiary did so after receiving a similar complaint.” The phrase did so might refer to the entire recorded-and-notified sequence or the more salient regulator-notification act. Legal or compliance stakes make that ambiguity important. A translator should not automatically conclude that the subsidiary completed both stages unless context supports that.
If the paragraph elsewhere details the subsidiary’s record and notice, the full sequence may be established. If it merely confirms notification, repeating recorded could invent a bookkeeping fact. Conversely, translating did so as recorded but omitting regulator notice could understate the reported action. High-stakes records need an explicit source-query workflow.
The safest response is to annotate candidate antecedents and request confirmation from an authorized source. A provisional bilingual draft can show a literal reference construction if natural in the target language, but avoid committing publicly to one unsupported reading. Legal review may also determine whether notification, submission and disclosure are distinct legally defined acts.
Do-so anaphora can be efficient English prose; that efficiency is not always compatible with precise regulatory traceability. When an important right, duty or deadline depends on the anaphor, transparency about ambiguity is part of accuracy, not a failure to translate.
Worked example: environmental monitoring and repeated methods
A field report states, “The first team sampled groundwater downstream of the plant. The second team did so upstream.” The repeated action is sampling groundwater. Upstream is new location information that contrasts with downstream. It would be wrong to translate “the second team sampled downstream as well,” because that erases the spatial contrast, or “the second team tested the plant,” because it changes the measured object.
The noun groundwater may refer to wells, aquifers or collection points depending on the methodology. Do not invent a specific sampling device, depth or chemical test. Sampling groundwater is not equivalent to finding contamination. The latter requires results and thresholds that the short sentence does not supply.
A precise expansion is “The second team sampled groundwater upstream.” This is an example where the target can be clearer by repeating the verb rather than imitating do so. Notice that repeating sampled does not obligate the translator to copy downstream. A mechanical copy of the entire first clause would contradict the new modifier upstream.
A practical review can list shared method, distinct locations, sample identity and results. Shared: groundwater sampling. Distinct: location relative to plant. Unknown: whether sampling depth, dates and analyses were the same. Preserving these boundaries helps a reader interpret a monitoring report without granting it results that have not been reported.
How do so differs from auxiliary do and from a short yes answer
“Nina reviews the data every day, and Omar does too” is ordinary VP ellipsis with auxiliary does supporting an understood predicate. “Nina reviews the data every day, and Omar does so too” uses the do-so expression. Both can refer to Omar reviewing data, yet the underlying linguistic constructions are not identical. Formal researchers debate how do-so anaphora is represented and which antecedents it can use; translator-relevant differences include register, event type and the role of surrounding modifiers.
“Did” can also be a lexical verb in sentences such as “They did the experiment.” There the word did names the action itself in a broad way. “They did so” is an anaphoric expression whose interpretation depends on previously established activity. “Yes, I did” is an answer to a prior question, often with its own auxiliary structure. Treating all three as one word-for-word target can create awkward or misleading language.
A beginner can learn a useful question: what precise activity would you insert if you were forbidden to use do, so, it or that? In a clear passage, the answer often exposes the core event. An advanced translator adds a second question: which part of the antecedent event is being referred to, especially when there are several predicates or modifiers?
For research context, compare the University of California dissertation on do-so anaphora and a Cambridge Journal of Linguistics paper on VP dependents and do-so. They show why the relation is more subtle than substituting the last verb.
When a new modifier narrows or changes the repeated event
“The first driver stopped at the marked line. The second did so just beyond it.” In a strict literal reading, stopping just beyond the line differs from stopping at the marked line. The phrase did so may be understood more generally as stopping the vehicle, while the new modifier revises the location. If a translator repeats the entire phrase stopped at the marked line and then adds just beyond it, the target becomes contradictory.
This tells us something important about anaphora: it often selects an event type at the appropriate level of generality, not a rigid textual string including every adjunct. The reader resolves which details are shared and which are contrasted. That resolution depends on discourse and world knowledge, but a translation should not add details beyond those available.
Another example: “Rita submitted the file electronically. Chen did so by hand at the service counter.” Depending on what by hand means, the shared event is submission, while the method changes. A target should keep both methods apart. It should not carry electronically into Chen’s submission as if it were an invariant.
To test the level of event reference, remove optional modifiers from the first clause until the repeated action and the second modifier can coexist naturally. This is a diagnostic, not license to delete important source details from the final translation. Keep each participant’s own modifiers in the right place while ensuring that the shared action is recognizable.
AI translation risks: fluent substitution, wrong antecedent
AI systems are good at expanding short anaphoric expressions into polished sentences, which makes an unnoticed wrong expansion particularly dangerous. A system may see a paragraph about checking, signing and sending a document, then decide that did so must mean sent it because sending is the expected final step. Unless the source context supports that, the model has supplied a plausible process story rather than a translation.
A robust AI-assisted review should request the candidate antecedent explicitly, in source-language words, before generating a target. Record any alternatives when more than one action is available. Then compare the target’s participant roles, timing, modality, object and new modifiers against the source. If an expansion adds a completed outcome where the original reports only intention, reject it.
Ask a second reviewer, human or model-assisted, to paraphrase the target without being shown the English expansion. Does the resulting event match the source’s understood action? If the target says approved, disclosed or delivered when the source only says reviewed, a problem will be visible. Reviewers should not defer to the system’s confidence or to a smooth style.
Sensitive records may require qualified human translators and domain specialists. The role of an AI system is to help identify possible readings and compare consistency, not to create authoritative legal, clinical or technical interpretations that the source record does not warrant.
Practice set: eight recoveries with explanations
1. “Lee filed the application, and Arun did so later.” Recover filed the application. Do not convert it to approved the application. 2. “The first group scanned all three slides, and the second did so with a different machine.” Recover scanning all three slides if that is the established antecedent; preserve the different machine and do not assume the image quality matched.
3. “The contractor agreed to inspect the roof and did so on Tuesday.” Here did so points to inspecting, not simply agreeing. Completion of an inspection is asserted, but successful repair is not. 4. “The director asked staff to secure the doors, and they did so.” Recover securing doors. If the context identifies which doors, preserve that set; do not turn securing into locking every door in the building without evidence.
5. “Students can request a transcript online. To do so, sign in to the portal.” Recover requesting a transcript online. The instruction is about that goal, not changing the transcript. 6. “Mina prepared the summary early; Paulo did so at the deadline.” Recover preparation and preserve the contrasting times.
7. “The clinic offered to share information, and the partner did so.” Do not automatically decide whether the partner offered or actually shared without checking how the earlier event is described; if the intended reading is unclear, query the writer. 8. “The editor corrected the table and updated its notes. The reviewer did so too.” Determine whether the second action includes both steps or one salient step; a risk-aware translation flags the scope rather than guessing.
For each exercise, create a short role map and one fully explicit paraphrase. Then rewrite the sentence naturally in the target language. Check whether another person can recover the intended action without being shown the source. The best solution is clear about what happened while restrained about what was not stated.
Frequently asked questions about did so and doing so
What does did so mean? It usually refers back to a previously described action or activity. The particular action depends on the discourse, not the dictionary entry for did. Is did so more formal than did it? In many contexts yes; Cambridge’s grammar guidance describes do so as relatively formal. Their exact acceptability and discourse effects need not be identical.
Can did so refer to multiple steps? It can refer to a complex discourse event, but the scope of that event must be inferred carefully. Do not assume every preceding action was repeated. Does did so always mean the same object was involved? No. Context determines whether the repeated action involved the same physical item, another item of the same type or another participant’s corresponding object.
Can I repeat the full verb when translating? Yes. Repetition is often the most readable way to preserve meaning in a target language without English-style do-so substitution. Does did so imply success? It implies performance of the action denoted, which may or may not involve a successful outcome depending on the verb and context. “Tried to repair” and “repaired” are not equivalent.
Why can “did too” sound better than “did so too” with know? Event type and the semantic properties of the anaphor matter, although there are documented complications and exceptions. Is did so the same as null complement anaphora? Not exactly; they are related ways of leaving content unspoken but have different forms and analyses.
What if two possible antecedents exist? Preserve ambiguity where appropriate or ask for clarification, especially in high-stakes records. How do I avoid adding information? Build an action-and-role grid, expand only source-supported content, and compare the final target with the original source.
Final accuracy checklist and learning route
Before publication, recover the exact event, not just the nearest verb. Check who performed it, when or how the second event differs, whether the source presents an actual event or merely a possibility, and which objects or conditions belong to each participant. If a new modifier conflicts with a detail in the first clause, do not mechanically copy the conflicting detail. Resolve the intended degree of shared activity.
Read the target-language sentence alone. Can an ordinary reader identify the repeated action? Does the result preserve the source’s level of specificity and formality? If a second reader infers approval where the source says review, a completed repair where the source says attempt, or the same method where the source says different method, revise it. Clear prose is valuable only when it reports the right event.
Continue through Master Art of Translation, Null Complement Anaphora and Ellipsis and Omitted Meaning. They cover related but distinct ways in which a reader recovers content that a sentence does not pronounce. The portable rule is simple: keep the event and its limits, even when the target language repeats the verb instead of saying did so.