资源 · 32
设计可访问的表单和可恢复的错误
引导用户输入、解释错误并保留用户完成任务的进度。
已更新 · 16 min
本指南有助于实现的目标
- 限制字段
- 命名和指南
- 解释错误
- 测试恢复功能
快速检查
- 每个字段都有易于理解的名称吗?
- 输入前是否解释了预期格式?
- 错误是否标识了其所在字段和可能的更正方法?
- 是否保留了有效值?
- 是否可以使用键盘和缩放完成任务?
分步指南
- 1
从任务开始
仅列出完成任务所需的数据以及用户必须理解的步骤。 删除与事务无关的请求。
交付物:表单和字段的最小说明。
- 2
标签和分组
为每个控件添加可见标签,将相关选项分组,并在附近放置帮助信息。 检查阅读顺序和 Tab 键顺序。
交付成果:已测试的结构和标签。
- 3
防止错误
在提交前提供格式示例和状态约束。 不要仅使用颜色来表示某个字段是必填项。
交付成果:说明和输入状态。
- 4
解释每个错误
验证后,发布摘要,将每个问题与其对应的字段关联,并描述如何解决。 保持正确的值和上下文。
交付成果:可纠正的错误场景。
- 5
处理故障
区分输入错误、业务拒绝、超时和服务器故障。 提供下一步操作,而不是假装请求已成功。
交付成果:每个状态的消息和恢复方案。
- 6
测试实际任务
使用键盘、缩放、移动设备和适当的辅助功能重现成功和失败的情况。 评估障碍,修复它们并重新测试。
交付成果:任务完成的证据。
管理指标
| 指标 | 衡量内容 | 首要行动 |
|---|---|---|
| 完整性 | 需要带有标签和帮助的字段 | 修复含义模糊的控件 |
| 错误 | 带有解释和字段链接的失败 | 重写消息 |
| 恢复 | 更正后保留值 | 修复状态管理 |
| 任务 | 使用键盘和缩放完成操作流程 | 消除障碍 |
常见错误
- 使用占位符代替标签
- 仅显示“发生错误”
- 错误后清除整个表单
- 仅使用鼠标测试正常流程
常见问题解答
占位符是否足够?
否。用户输入值时,应始终显示可识别的标签。
错误信息应何时显示?
选择一个合适的时机,并在不造成不必要干扰的情况下解释错误原因和解决方法。
如果服务器发生故障怎么办?
不要声称成功;保留可以保存的内容,并提供易于理解的恢复方案。
表单:准则、错误与对比度
此语言版本的详细参考说明以英语提供。
An accessible form lets people understand the request, reach controls, correct an error and finish the task. For an incident-reporting portal, service request or authentication journey, quality also depends on waiting, failure and backtracking states.
This editorial selection connects 17 WCAG 2.2 Level A or AA criteria to review questions and evidence to retain. Identifiers, English titles and levels were checked against the W3C Recommendation of 12 December 2024, consulted on 23 October 2026. The selection does not cover every criterion applicable to a page or process.
Separate the rule, its explanation and observed evidence
| Reading layer | What it provides | How to use it |
|---|---|---|
| WCAG 2.2 Recommendation | Normative text: read the criterion, definitions and exceptions together. | Connect the finding to the exact criterion and examined scope. |
| WAI Understanding and tutorials | Informative explanations and guidance for interpretation and implementation. | Use them to prepare checks without presenting them as new normative requirements. |
| Journey observation | A check on a page, version, state and configuration. | Retain reproduction steps and evidence; a calculation or scanner does not replace complete evaluation. |
W3C · Web Content Accessibility Guidelines (WCAG) 2.2 · Recommendation · 12 December 2024
17 criteria for a form review
Questions and checks below are CIE editorial methodology. Each criterion links to the dated normative text; its Understanding link leads to an informative explanation. Levels are neither a score awarded to this site nor an audit result.
| W3C criterion and level | CIE review question | Evidence or check to prepare |
|---|---|---|
| 1.3.1 · Info and Relationships Level A | Do field labels and groups remain understandable in the semantic structure? | Accessibility tree and traversal of named groups. |
| 1.3.5 · Identify Input Purpose Level AA | For personal inputs covered by the taxonomy, can software identify the input purpose? | Purpose-appropriate autocomplete tokens; distinguish a label from a purpose token. |
| 1.4.1 · Use of Color Level A | Can an error or selection be understood without relying on colour alone? | Error text, field identification and state independent of red or green. |
| 1.4.3 · Contrast (Minimum) Level AA | Does meaningful text stand out from its background in relevant states? | Actually adjacent colours, text size and weight; criterion threshold and exceptions. |
| 1.4.11 · Non-text Contrast Level AA | Is visual information needed to identify components and states distinguishable? | Necessary border, icon or other cue compared with adjacent colours; examine exceptions. |
| 2.1.1 · Keyboard Level A | Can the journey’s actions be operated by keyboard within the criterion’s scope? | Pointer-free scenario: selection, entry, help, correction and submission. |
| 2.4.7 · Focus Visible Level AA | Can visitors see which interactive element has keyboard focus? | Tab and Shift+Tab through initial, error and confirmation states. |
| 2.4.11 · Focus Not Obscured (Minimum) Level AA | Does the focused component remain at least partly visible despite author-created content? | Test sticky headers, banners and panels; this AA minimum does not mean no obscuration at all. |
| 2.5.3 · Label in Name Level A | Does the accessible name contain visible label text where this criterion applies? | Compare visible text and computed name, including for voice control. |
| 2.5.8 · Target Size (Minimum) Level AA | Do pointer targets reach 24 × 24 CSS pixels or satisfy a criterion exception? | Measure bounds and examine spacing, equivalent control, inline text, user-agent control or essential presentation. |
| 3.3.1 · Error Identification Level A | When an input error is automatically detected, are its field and nature described in text? | Submit a missing value and an incorrect format; retain the message associated with the field. |
| 3.3.2 · Labels or Instructions Level A | Are necessary labels or instructions available before entry? | Requirements known in advance: expected field, mandatory entry and useful format. |
| 3.3.3 · Error Suggestion Level AA | Is a known correction suggested where it does not compromise security or content purpose? | Compare a generic message with actionable help for the error actually detected. |
| 3.3.7 · Redundant Entry Level A | Within one process, is previously supplied information populated or available to select, subject to exceptions? | Return to a step; distinguish still-valid data, essential requirements and security needs. |
| 3.3.8 · Accessible Authentication (Minimum) Level AA | Does an authentication step require a cognitive test without an alternative, assistance or specified exception? | Usable password-manager, paste and other mechanisms; read the complete exceptions. |
| 4.1.2 · Name, Role, Value Level A | Are component name, role, value and relevant states exposed to assistive technologies? | Inspect native and custom controls, then change their states. |
| 4.1.3 · Status Messages Level AA | Can relevant status messages be presented by assistive technology without receiving focus? | Waiting, error or brief result; test announcements and avoid unnecessary interruptions. |
Five recovery situations to write and test
| Illustrative situation | Useful feedback | Check |
|---|---|---|
| Required field left blank | Identify the field and explain the expected information in text. | Keep the message associated with the field and the label visible. |
| Unaccepted format | Explain the actual accepted format and suggest a correction where possible. | Do not erase other valid entries; let the person resume correction. |
| Multiple input errors | Provide an overview and messages close to the fields. | Error-summary links reach the correct controls and keyboard order remains coherent. |
| Service waiting or failure | Say whether the request is pending, failed or requires another action. | Useful status feedback is perceivable without repeated announcements on every keystroke. |
| Returning to a step in the same process | Retain still-valid information or make it available to select. | Examine 3.3.7 exceptions and avoid unjustified forced re-entry. |
W3C WAI · Forms Tutorial · Labeling Controls · Practical guidance · consulted 2026-10-23
W3C WAI · Forms Tutorial · User Notifications · Practical guidance · consulted 2026-10-23
本地对比度计算器
Calculation uses two opaque, solid sRGB colours. Enter #RGB or #RRGGBB, case-insensitively. The ratio uses relative luminance and (lighter L + 0.05) / (darker L + 0.05). Numeric display is rounded for reading; threshold comparisons use the unrounded result.
This model does not inspect website styles and does not handle transparency, gradients, background images, video, other colour spaces or rendering variations. It compares numbers with thresholds; it does not decide whether a criterion applies or its exceptions are satisfied.
Calculated ratio : ·
| Reference and use | Ratio threshold | Comparison for the entered pair |
|---|---|---|
| 1.4.3 · AA · normal text | 4.5:1 | |
| 1.4.3 · AA · large text | 3:1 | |
| 1.4.6 · AAA · normal text | 7:1 | |
| 1.4.6 · AAA · large text | 4.5:1 | |
| 1.4.11 · AA · relevant non-text visual information | 3:1 |
WCAG large-scale text includes at least 18 point, or 14 point bold, or equivalent sizing under the definition’s rules. Text merely looking “large” on screen does not justify choosing a reduced threshold. For non-text, examine necessary identifying information and all relevant adjacent colours.
W3C WAI · Understanding SC 1.4.3 · Contrast (Minimum) · Informative explanation · consulted 2026-10-23
W3C WAI · Understanding SC 1.4.6 · Contrast (Enhanced) · Informative explanation · consulted 2026-10-23
W3C WAI · Understanding SC 1.4.11 · Non-text Contrast · Informative explanation · consulted 2026-10-23
Seven calculated pairs, available without JavaScript
| Learning case | Foreground / background | Calculated ratio | Numeric thresholds reached |
|---|---|---|---|
| Illustrative gold and dark background | #D4BD7F / #101719 | 9.826201:1 | 3:1 · 4.5:1 · 7:1 |
| Black and white | #000000 / #FFFFFF | 21.000000:1 | 3:1 · 4.5:1 · 7:1 |
| Grey below the 4.5 threshold | #777777 / #FFFFFF | 4.478089:1 | 3:1 |
| Grey just below the 3 threshold | #959595 / #FFFFFF | 2.995346:1 | Numeric threshold not reached |
| Grey above the 7 threshold | #595959 / #FFFFFF | 7.004729:1 | 3:1 · 4.5:1 · 7:1 |
| Red and white | #FF0000 / #FFFFFF | 3.998477:1 | 3:1 |
| Identical colours | #777777 / #777777 | 1.000000:1 | Numeric threshold not reached |
#777777 grey on white gives approximately 4.478089:1: the 4.5:1 threshold is not reached. #959595 on white gives approximately 2.995346:1: rounding it to 3.0 does not make it reach the 3:1 threshold. These are calculations, not measurements of a real journey.
A task-oriented verification method
- Define the task
Specify what the person needs to complete and the pages and states involved: entry, waiting, error, correction and confirmation.
- Test different uses
Navigate by keyboard, with zoom and on a small screen; also inspect names and states exposed to assistive technology.
- Keep reproducible findings
Record version, configuration, steps, expected and observed results. Link barriers to criteria and retain the exceptions examined.
- Retest after correction
Recheck the whole task with the actual components and services, retaining any points still requiring confirmation.
Copyright © 2020–2024 World Wide Web Consortium. W3C Recommendation WCAG 2.2, 12 December 2024. Copyright © 2023 W3C®. This document includes material copied from or derived from Web Content Accessibility Guidelines (WCAG) 2.2. Identifiers, titles and levels are attributed to W3C; questions, examples and method are CIE synthesis. This is not an official translation or a modified technical specification.
W3C · Web Content Accessibility Guidelines (WCAG) 2.2 · Recommendation · 12 December 2024
W3C · Document license · 2023 version · Terms and attribution consulted 2026-10-23
Verified documentary edition
WCAG 2.2 · Recommendation · 2024-12-12 · 2026-10-23
Source HTML SHA-256: 6e3c5fe397257cae509a2fb4752b73062cf8cbeb92c2cec618989b17e4cf7057
官方参考文献
参考文献支持该方法。 请根据您的实际情况调整检查;它们并非认证。 原始参考文献标题和源文档可能使用其他语言。
参考资料查阅日期 .
- W3C WAI — Forms tutorial
- W3C WAI — Form validation
- W3C · Web Content Accessibility Guidelines (WCAG) 2.2
- W3C WAI · Understanding SC 1.4.3 · Contrast (Minimum)
- W3C WAI · Understanding SC 1.4.6 · Contrast (Enhanced)
- W3C WAI · Understanding SC 1.4.11 · Non-text Contrast
- W3C WAI · Forms Tutorial · Labeling Controls
- W3C WAI · Forms Tutorial · User Notifications
- W3C WAI · Understanding SC 3.3.8 · Accessible Authentication (Minimum)
- W3C · Document license · 2023 version
