Word Prediction Software¶

Strategy · All strategies
Evidence · 3 claims (2 for, 1 mixed) · 7 studies (3 review, 2 causal, 1 quant-synthesis, 1 theoretical), q1–q4 · 1 of 7 report an effect size

Description¶

Word prediction software displays a list of likely next words as a learner types, typically triggered by the first one or two letters. The learner selects a predicted word rather than typing it in full, which reduces spelling demands, keystrokes, and the working-memory burden of holding a word in mind while transcribing it. Most modern tools combine prediction with text-to-speech feedback, custom dictionaries, and topic-specific vocabulary.

Design Implications¶

Word prediction functions as an accommodation that offloads low-level transcription so working memory is available for planning and composing Chunking reduces working memory load. [+M]. Because transcription fluency is a bottleneck for struggling writers, freeing it can improve both writing volume and quality Automatic word recognition frees resources for comprehension. [+M]. The trade-off is that scanning and evaluating prediction lists imposes its own cognitive load; poorly tuned prediction can interrupt the flow of composition Cognitive overload degrades learning. [~M].

Context¶

Requirements¶

• A prediction engine tuned to the learner's vocabulary and topic (custom dictionaries, topic word lists)
• Prediction lists short enough to scan quickly (typically 3–6 items) with keyboard or single-switch selection
• Text-to-speech readback so learners can verify that the selected word is the intended one
• Explicit instruction in how to use the tool — learners who ignore predictions gain little

Constraints¶

• Prediction interrupts composition for fluent spellers and fast typists; for them it slows writing rather than helping [-M]
• Learners must be able to recognize the correct word in the list; prediction does not help if the target word is not already in the learner's receptive vocabulary [-M]
• Heavy reliance on prediction without spelling instruction can leave underlying encoding skills underdeveloped [~W]
• Poorly matched prediction (wrong frequency, wrong topic) adds scanning and decision costs that exceed the keystrokes saved [~M]

Implementation Variability¶

• Phonetic prediction: suggests words matching the learner's invented spellings (e.g., typing "fone" surfaces "phone")
• Topic-specific dictionaries loaded per assignment to raise hit rates in content-area writing
• Combined tools pairing prediction with speech recognition for learners whose oral language exceeds their written output
• Abbreviation expansion as a lower-cost alternative for recurring phrases

Target Learners¶

• Students with dyslexia, dysgraphia, or other spelling-related disabilities, for whom transcription is the primary writing bottleneck [+M]
• Students with physical disabilities who benefit from reduced keystrokes [+M]
• Emergent writers and English learners, where recognition-based selection supports vocabulary use [~W]
• Fluent spellers and fast typists, who gain little and may be slowed [-M]

Target Learning Goals¶

• Written expression: increasing length, fluency, and organization of compositions
• Spelling accuracy in produced text (distinct from spelling knowledge)
• Writing stamina and willingness to revise, by lowering the cost of producing text

Instructions¶

1. Assess the learner's transcription bottleneck: confirm that spelling/keystroke demands, not idea generation, are the limiting factor.
2. Configure the tool — set prediction list length, enable phonetic matching, and load topic dictionaries for the current assignment.
3. Model tool use explicitly: demonstrate scanning the list, listening to readback, and selecting, using a think-aloud approach.
4. Pair the tool with practice on authentic writing tasks so prediction becomes automatic rather than attention-demanding.
5. Fade support as fluency develops — reduce prediction frequency or move to abbreviation expansion — and continue targeted spelling instruction alongside.

Related Strategies¶

• Text-to-Speech — complementary readback that lets learners verify predicted words
• Speech-to-Text — alternative transcription accommodation for learners with stronger oral than written language
• Scaffolded Writing Frames — addresses planning and structure rather than transcription

Examples¶

• Co:Writer (Don Johnston) — widely used in special education; offers phonetic prediction and topic dictionaries; https://learningtools.donjohnston.com/product/cowriter/
• Clicker (Crick Software) — word prediction plus grids for elementary writers; https://www.cricksoft.com/us/clicker
• Read&Write (Texthelp) — prediction bundled with literacy supports in mainstream classrooms; https://www.texthelp.com/products/read-and-write/
• iOS/Android built-in keyboards — consumer-grade prediction now standard, though not configurable for instructional purposes

Key Sources¶

• MacArthur, C. A. (2009). Reflections on research on writing and technology for struggling writers. Learning Disabilities Research & Practice, 24(2), 93–102. doi:10.1111/j.1540-5826.2009.00283.x
• Evmenova, A. S., Graff, H. J., Jerome, M. K., & Behrmann, M. M. (2010). Word prediction programs with phonetic spelling support: Performance comparisons and impact on journal writing for students with writing difficulties. Journal of Special Education Technology, 25(4), 33–48.
• Graham, S., & Perin, D. (2007). A meta-analysis of writing instruction for adolescent students. Journal of Educational Psychology, 99(3), 445–476. doi:10.1037/0022-0663.99.3.445
• Berninger, V. W., & Winn, W. D. (2006). Implications of advancements in brain research and technology for writing development, writing instruction, and educational evolution. In C. A. MacArthur, S. Graham, & J. Fitzgerald (Eds.), Handbook of Writing Research (pp. 96–114). Guilford Press.