ALTA Competency Exam (CALT) common mistakes
ALTA Competency Exam (CALT) · Understand the traps and how to avoid them.
Translating long source sentences clause-by-clause without first mapping their structure, so a relative clause or prepositional phrase gets attached to the wrong noun (e.g., a modifying phrase describing the contract rather than the party).
Candidates read for general meaning and start drafting mid-sentence; in legal and administrative prose the modifier sits several lines from its head noun, so short-term memory, not analysis, decides the attachment.
What to do instead
Before translating, bracket the source into clauses, mark each subject-verb-object core, and draw an arrow from every relative clause and prepositional phrase to its head noun. Only start drafting once the attachment is written down.
Skipping or guessing at a low-frequency legal or administrative term (a procedural Latinism, an institutional title, a jurisdiction-specific office) and translating it as if it were the ordinary word it resembles.
The candidate has a working general vocabulary but no domain glossary, and under time pressure an approximate guess feels safer than pausing.
What to do instead
Keep a running list of every source term you cannot define in one sentence and resolve each one from a monolingual legal dictionary or a parallel text in the same jurisdiction before drafting; if still unresolved, render it neutrally and flag it rather than guessing.
Dropping or compressing elements during transfer: one item disappears from a list of three, a negative is lost, or a qualifier such as 'unless otherwise agreed in writing' is silently omitted.
Enumeration-heavy boilerplate feels repetitive, so the eye skips; negations and hedges are short unstressed words that do not survive gist-level reading.
What to do instead
Do a second pass in which you count enumerated items in source and target, and tick off every negation, modal, and conditional against a fixed checklist (not, no, unless, provided that, subject to, except).
Flattening modality: rendering 'shall', 'may', 'must not', and 'is entitled to' with the same generic target verb or with a modal that carries the wrong strength, turning a permission into an obligation.
Target-language modal systems rarely map one-to-one onto English legal modals, and the candidate translates by feel instead of by function.
What to do instead
Build a personal modality table before the exam mapping each source modal to a target equivalent plus its strength (obligation / permission / prohibition / discretion), and verify every modal instance against it in review.
Producing target sentences that reproduce source word order and noun-phrase length, yielding heavy premodification, stacked passives, and clauses that no native writer of the target would produce.
The candidate drafts literally and then only checks for meaning, never re-reading the target as an independent text in its own language.
What to do instead
After the literal draft, do a target-only rewrite pass with the source covered: split overlong sentences, move the verb to its natural position, and convert passive to active where the target prefers it. Read each sentence aloud to catch what the eye accepts.
Mismatching register: using informal address forms or everyday vocabulary in a formal certificate or court document, or the reverse - importing archaic legalese into a plain-language administrative notice.
The candidate focuses on propositional content and treats register as decoration, and does not notice that forms of address (tu/vous, tú/usted, du/Sie) encode legal relationships.
What to do instead
Identify register markers in the source on the first read (address forms, honorifics, formulaic openings and closings, nominalization density) and fix the corresponding target conventions before drafting the first sentence.
Rendering the same source term differently across one document - three translations for the same institutional title or defined term - or falling for a false cognate that is plausible in general language but wrong in the legal field.
Translation happens sentence by sentence with no memory aid, and cognates are accepted because they read fluently rather than because they are correct in the domain.
What to do instead
Build a glossary during the initial read-through covering defined terms, titles, and recurring phrases, then apply it consistently and run a search on each term at the end to confirm one rendering per concept.
Leaving source-language mechanics in the target: unconverted date and number formats, wrong decimal and thousands separators, missing or misplaced diacritics, and source capitalization carried into headings and titles.
These items are processed as formatting rather than content, so they fall outside the candidate's attention during translation and are not covered by a meaning-focused review.
What to do instead
Add a mechanics checklist to the final review: convert every date, figure, currency, and percentage to target conventions, verify diacritics and agreement on every edited word, and check heading and title capitalization against target norms.
Leaving fragments of the source language in the target text, or introducing agreement and spelling errors during self-editing - a verb re-conjugated but its subject left in the old number, an article left over from a deleted phrase.
Revision is done on screen within the same session, where the brain reads what was intended rather than what is written; short source fragments sitting in the target look 'finished' to the tired eye.
What to do instead
Run a dedicated monolingual proofread pass in the target only, and use a mechanical check for leftover source tokens (search for common source-language function words) plus a read-through from the last sentence backwards to break the expectation effect.
Improving, explaining, or softening the source - adding a clarifying phrase, tidying an awkward original, or resolving an ambiguity silently - instead of translating it and flagging the problem separately.
The candidate equates a good translation with a smooth target text and believes an unexplained ambiguity would look like a mistake, so the fix is hidden inside the translation.
What to do instead
Translate exactly what the source says, then record any ambiguity, illegible segment, or apparent source error in a translator's note or query to the client; treat a documented question as the professional deliverable, not a defect.