Assessment Accommodations and Score Interpretation: A Guide for Employers

A candidate completes an online reasoning assessment with additional time approved in advance. Another uses a screen reader to access the same questions. A third takes an equivalent-format work sample because the original interface was inaccessible. All three receive scores. Can the employer compare those numbers directly with the scores of candidates who completed the standard assessment?

Sometimes there is a defensible basis for comparison; sometimes there is not enough evidence. The presence of an accommodation does not automatically invalidate a score, but neither does an identical scoring scale guarantee that the scores mean the same thing. Employers must distinguish an accommodation that removes an irrelevant barrier from a modification that may alter the ability being measured.

This guide addresses the interpretation and decision stage after an accommodation has been approved. For how to receive and decide requests, see Reasonable Adjustments in Pre-Employment Testing. For general platform accessibility, see Accessible Online Assessments. Those topics are important, but the central question here is what a resulting score can support.

Why an accommodated score requires a clear interpretation

Every assessment score is an observation made under particular conditions. A score of 32 out of 40 does not independently reveal which abilities produced the result, how stable it would be on another occasion or whether the testing conditions matched those of other candidates.

An employer may intend the score to represent numerical reasoning, written communication, spreadsheet skill or situational judgement. The test conditions should allow the candidate to demonstrate that construct while minimising barriers unrelated to it. An accommodation may improve the validity of the intended interpretation by removing such a barrier. However, it may also affect timing, task demands, response processes or the meaning of a norm-based comparison.

Three questions guide responsible use:

1. Construct: What was the assessment supposed to measure?
2. Change: What exactly differed from the standard administration?
3. Evidence: What supports the proposed interpretation and comparison under the changed conditions?

These questions are distinct from whether the candidate was entitled to an accommodation under applicable law. Employers should fulfil their obligations without making unsupported psychometric claims, and should obtain legal and specialist advice where needed.

Accommodation, accessibility feature or different assessment?

Not every difference in administration is the same kind of change.

A useful classification should be recorded before results are examined. Otherwise, an employer may rationalise an accommodation as “equivalent” because the preferred candidate scored well—or as “invalid” because the candidate scored poorly.

Step 1: Revisit the original score claim

A score report should state what the assessment was designed to support. Consider two tests that each last 30 minutes.

Test A: A numerical-reasoning exercise where the employer wants to know whether the candidate can interpret figures accurately. The time limit was set mainly to fit the recruitment schedule.

Test B: A simulation of a safety-critical monitoring task where rapid detection and response are demonstrated job requirements. Time pressure is part of the work being represented.

An extension to Test A may remove an incidental obstacle without changing the core ability of interest. An extension to Test B could alter the very performance requirement being tested. That does not mean the candidate must be denied access to an appropriate assessment; it means the employer needs a suitable method for evaluating the essential competency and must not assume the same raw score has the same interpretation.

The employer should retain the job-analysis rationale, blueprint, test specification and any validation evidence. A vendor's default timer is not proof that speed is essential.

Step 2: Describe the change precisely

“Accommodation provided” is too vague for technical review. An assessment specialist may need to know whether the change involved presentation, timing, response mode, breaks, environment, instructions or content. At the same time, hiring managers generally do not need a candidate's diagnosis or sensitive supporting documents.

Use a restricted technical record containing only what is necessary for legitimate evaluation. For example:

A separate, access-controlled record may hold information required to administer the accommodation. Avoid copying medical details into the scoring file or ordinary hiring notes. Retention, disclosure and lawful basis for processing must follow the relevant jurisdiction's rules.

Step 3: Distinguish raw-score equality from measurement equivalence

Suppose Candidate A and Candidate B each answer 34 of 40 questions correctly. The scores are numerically identical: 34/40 = 85%. This does not establish that both candidates were assessed under conditions with the same measurement properties.

If the accommodation only changed an inaccessible control while preserving the item content and response process, the comparison may be reasonably supportable, depending on the evidence. If one candidate received unlimited time on a task where speed contributes materially to the score, the interpretation may differ. If one candidate completed a substantially different task, equal percentages may have no meaningful cross-form relationship.

A raw percentage is not an equating study. Converting two different forms to a 0–100 scale does not make them interchangeable. Equating, linking or comparability analysis requires a defensible design and suitable evidence. Employers should avoid informal “corrections” that add or subtract points based on an accommodation label.

Step 4: Consider the evidence hierarchy

The amount of evidence needed should reflect the stakes of the decision, the degree of modification and the consequences of an error. The following is a practical planning hierarchy, not a universally mandated statistical standard.

Design evidence: A documented construct definition and expert analysis of which task demands are essential. This is necessary but rarely sufficient by itself.

Response-process evidence: Observations or interviews showing how candidates actually understand and complete the task under each condition. These can reveal barriers or changed strategies.

Item and form evidence: Reviews of item content, scoring keys, presentation, timing and accessibility. Where suitable data exist, examine item functioning, score distributions and reliability.

Comparability evidence: Empirical analyses designed to evaluate whether scores support the intended cross-condition comparisons. Methods depend on the assessment, sample and question.

Decision evidence: Monitoring whether the assessment supports relevant workplace outcomes and whether its use produces unexpected consequences. Correlation with job performance can be informative but must be interpreted carefully.

Small pilots may identify obvious problems yet be too limited to establish equivalence. The correct response to thin evidence is a qualified interpretation or an alternative decision method—not a fabricated statistical adjustment.

Worked example: additional time and score meaning

An employer uses a 40-item data-interpretation test. Its scoring system awards one point per correct answer and no penalty for wrong answers. The employer has not demonstrated that finishing speed is an essential part of the role.

Candidate P completes the standard 45-minute administration and answers 31 items correctly. Candidate Q receives an approved 60-minute administration and answers 33 correctly.

The arithmetic is simple: 31 ÷ 40 × 100 = 77.5%; 33 ÷ 40 × 100 = 82.5%. But the five-percentage-point difference is not automatically a five-point difference in underlying ability. Nor can the employer assume that Q's score is inflated by a particular number of points.

A careful interpretation might be: “Candidate Q demonstrated 33 correct responses under the approved conditions. The assessment has not yet established full comparability of completion-time effects across these conditions. Use the result alongside other job-related evidence and seek specialist review before applying a speed-sensitive rank or cutoff.”

If the intended construct is accuracy and timing was administrative, the accommodated result may be highly relevant. If rapid throughput is an essential construct, the employer may need a separate, accessible work simulation of that demand. The appropriate choice is based on evidence, not on a universal rule about extra time.

Worked example: alternative-format task

A candidate cannot access a visual dashboard exercise. The employer offers an accessible data table with equivalent underlying figures and asks the candidate to identify the same operational anomalies.

This alternative could preserve the analytical target while changing the visual-reading demands. That is appropriate only if visual dashboard interpretation is not itself an essential competency. The employer should review whether the revised format changes cognitive load, available cues or difficulty.

If the original form awards 20 points and the alternative awards 20 points, the employer should not assume that a score of 16 on each form represents identical performance. It can evaluate the same competencies using a common rubric and, if sufficient evidence is lacking for numeric linking, report the demonstrated behaviours separately. A hiring panel might use structured work-sample evidence and a job-related interview rather than forcing an unsupported direct ranking.

The important distinction is between fair consideration of evidence and unjustified score equivalence. An employer can make a defensible decision without claiming more precision than the assessment supports.

Norms, percentiles and pass marks need particular care

A percentile ranks a score relative to a defined reference group. It is not a percentage of questions answered correctly. If the norm group completed the assessment under different conditions, the meaning of an accommodated candidate's percentile may require further evidence.

For example, a raw score of 32 might correspond to the 75th percentile in a particular standard-administration reference sample. It does not follow that 32 under every modified administration is also at the 75th percentile of the same construct. The norm's applicability is a technical question, not a formatting choice.

Similarly, a pass mark should be tied to the intended competencies and decision rules. A cut score established under one administration may or may not transfer to another. Employers should consult relevant assessment specialists before treating thresholds as interchangeable. Do not lower, raise or waive a standard ad hoc because an accommodation was used, and do not exclude a candidate simply because the evidence needs careful interpretation.

When uncertainty is material, use a pre-authorised process for additional job-related evidence rather than quietly manipulating the score.

Do not penalise candidates for using approved arrangements

The purpose of an accommodation is to address a barrier, not to create a hidden negative indicator. A score report should not contain a warning such as “candidate required extra time—reduce confidence” merely because the candidate used an approved arrangement.

Technical limits on a particular score comparison can be documented without implying that the candidate is less capable. Separate three ideas:

• Administration fact: A specified, approved condition was used.
• Measurement limitation: Particular inferences may lack adequate supporting evidence.
• Hiring decision: The employer must evaluate job-related evidence fairly under applicable policy and law.

Only authorised personnel should access the minimum accommodation information needed for their function. A hiring manager may need a qualified score interpretation or alternative evidence, not the underlying personal circumstances.

Build a decision rule before a difficult case arrives

A transparent score-use policy reduces inconsistent treatment. Consider four interpretation categories:

These categories are a governance aid, not a validated classification system. A competent assessment lead should decide the category with relevant evidence and record the rationale. Candidates should have access to a clear route for raising concerns about administration and score use.

Example decision memo

Assessment purpose: Evaluate reconciliation accuracy for an accounts officer role.

Condition: Candidate completed the same item set with an accessible input method.

Evidence: Subject-matter experts confirmed that mouse dexterity is not a job requirement; item wording and scoring keys were unchanged; accessibility testing confirmed correct recording of responses.

Interpretation: The observed correct-response count supports an inference about reconciliation accuracy under the approved presentation. Any remaining uncertainty about time effects should be documented rather than assumed away.

Decision: Use the work-sample result under the established rubric, with structured interview evidence as originally planned. No negative notation about the candidate's adjustment appears in the hiring panel report.

This memo illustrates a defensible reasoning process; it is not a legal determination or proof of statistical equivalence.

What a hiring manager should see

A well-designed report distinguishes observations from conclusions. For example:

Observed performance: 33 of 40 data-interpretation questions answered correctly. Intended interpretation: Evidence of numerical interpretation accuracy under the administered conditions. Decision guidance: Consider with the structured work sample and other approved criteria. Direct speed-based ranking is not supported by the evidence currently available. Confidentiality: No personal accommodation details are included in this report.

Do not expose a medical condition, request history or unsupported “adjustment penalty” in a recruiter dashboard. If the employer must retain an administration flag for audit, control who can see it and why.

Review evidence without overclaiming statistics

Employers may be tempted to compare the average scores of accommodated and non-accommodated candidates. Such a comparison is easily misread. The groups may differ in role, experience, form, construct, timing and many other factors. A mean difference alone does not establish that an accommodation helped or harmed performance.

A specialist might examine completion-time distributions, item-level patterns, reliability, response processes and differential functioning. More sophisticated designs may evaluate measurement invariance or predictive relationships. These analyses need adequate samples, appropriate assumptions and careful interpretation.

In a small programme, qualitative evidence and expert review may be more realistic than elaborate modelling. That can support limited operational decisions, but the employer must state what remains unknown. “No evidence of a difference” is not the same as “evidence of equivalence.”

Special cases that deserve explicit rules

Interrupted tests and paused timers

If an authorised break pauses a timer, record whether the candidate could view questions during the break. A paused clock with questions visible may create a different opportunity from a fully locked break. The score-use review should reflect the actual implementation, not just the label “rest break.”

Screen readers and complex diagrams

A screen reader can remove an interface barrier, but a text description of a diagram may alter the task. Determine whether visual extraction is essential to the job or whether the intended construct is interpretation of the underlying information. Review the accessible alternative with relevant experts.

Human readers or scribes

A reader's intonation, clarification or unintended cues may influence responses. A scribe may introduce transcription errors or change response speed. Clear protocols, training and quality checks are needed; equivalence cannot be assumed merely because the content appears unchanged.

Translation or bilingual administration

An approved translated form raises questions about language adaptation, cultural meaning and score linking. This is a distinct technical problem from a timing adjustment. See Language Proficiency Versus Job Skills in Assessment Design for the construct-definition stage.

Retests after technical failure

A retest may involve prior exposure to items, changed practice effects or a different form. Apply the incident and retake policies consistently, preserving evidence of what occurred. Do not label a technical retest as an accommodation without a valid reason.

A practical implementation workflow

Before launch: Define the intended construct, permissible standard features, approved accommodation routes, scoring rules and the person authorised to decide comparability questions. Specify how alternative evidence will be used if direct comparison is unsupported.

At request approval: Record the functional change needed and evaluate whether it is expected to affect task content, timing or response processes. Keep sensitive information in the appropriate restricted system.

Before administration: Verify that the intended condition actually works. Test timer behaviour, accessibility, response recording and scoring. Confirm the candidate receives accurate instructions.

After administration: Check the actual administration log, score calculation and any deviations. Do not assume that an approved setting was implemented correctly.

Before hiring decisions: Apply the pre-established interpretation category and report limitations in plain language. Where a score is not directly comparable, use an authorised alternative rather than a made-up conversion.

After the decision: Review incidents, complaints and evidence gaps. Update policy prospectively through change control; avoid retrospective rules invented for a particular candidate.

A quality-control checklist for assessment leads

• Is the intended construct documented and job-related?
• Does the record specify what changed, without unnecessary sensitive details?
• Did the candidate complete the test under the approved conditions?
• Are raw-score calculations correct?
• Is there evidence for the intended comparison, or is the limitation explicit?
• Are norms, percentiles and pass marks used only where justified?
• Does the hiring report avoid penalising accommodation use?
• Is any alternative decision method pre-authorised and consistently applied?
• Are score records, access permissions and retention controls appropriate?
• Is there a candidate route to query an error or request review?

Frequently asked questions

Does extra time always make an assessment score incomparable?

No. The effect depends on what the test measures, why it is timed and the available evidence. Extra time can remove an irrelevant barrier in some assessments; in others, timing is part of the target competency. Neither automatic equivalence nor automatic invalidity is justified.

Should a candidate's accommodated score be reduced?

There is no general basis for subtracting points because an accommodation was used. Any score transformation would require a specific defensible measurement rationale and appropriate evidence. An arbitrary penalty undermines fairness and may create legal risk.

Can a hiring manager see that an accommodation was used?

Access should be limited to what is necessary for a legitimate decision or operational responsibility and permitted by applicable law. Usually the manager needs an appropriate interpretation and any decision-relevant limitations, not confidential personal information.

What if a vendor claims its accommodated scores are equivalent?

Ask for evidence specific to the accommodation, assessment, population and proposed use. A general accessibility statement or the fact that the scoring formula is unchanged does not establish equivalence. Document gaps and seek specialist review for high-stakes decisions.

Can two different assessment forms be compared if both use percentages?

Not on that basis alone. Equal percentages can represent different difficulty and measurement properties. Direct comparisons require a supported linking or equating rationale, or a decision framework that does not rely on numeric interchangeability.

Is it unfair to use an alternative work sample when direct score comparison is uncertain?

Not necessarily. A well-designed, job-related alternative can provide meaningful evidence when the original form is inaccessible or not interpretable for the intended purpose. The method, rubric and decision rules should be approved, proportionate and consistent with applicable obligations.

How should an employer handle an uncertain result near the pass mark?

Do not invent a score adjustment or silently reject the candidate. Apply the documented uncertainty procedure: technical review, verification of administration, consideration of other approved job-related evidence or a suitable alternative assessment where appropriate. Explain the process without disclosing confidential information.

Conclusion: interpret what the assessment actually showed

Accommodations can make assessment evidence more meaningful by removing barriers unrelated to the job. But a fair process also requires disciplined score interpretation. The employer must know what changed, what stayed constant and which comparisons are supported by evidence.

The goal is neither to treat every administration as identical nor to distrust every accommodated result. It is to make job-related decisions that are transparent about what the evidence can—and cannot—establish. For organisations using Talentrenza Assess, the same principle applies: configure and administer assessments thoughtfully, then interpret results in light of the role, test conditions and available validity evidence. No platform setting alone establishes score equivalence.

Professional references

• Society for Industrial and Organizational Psychology. (2018). Principles for the Validation and Use of Personnel Selection Procedures. Publisher record.
• International Test Commission. Guidelines for Translating and Adapting Tests, second edition. Official guidance.
• World Wide Web Consortium. Web Content Accessibility Guidelines (WCAG) 2.2. Official standard.