What can we prepare before speaking?

Roelofs (2000)

Utku Turk

LING 449U, How words are built

2026-09-22

The conditions: cue, response, shared part

Each row is a three-pair block. Learn cue → response; then see only the cue and say its response.

Cue	Response	Condition name	Shared
religie haast spoor	bijbel bijna bijster	Syllable: shared	Initial syllable bij
religie brein dicht	bijbel hersens nader	Syllable: mixed	No common beginning
studie toneel long	bijvak bijrol bijnier	Morpheme: shared	Initial morpheme bij
toneel steiger mond	bijrol herbouw nasmaak	Morpheme: mixed	No common beginning

religie → bijbel: religion → bible
toneel → bijrol: stage → supporting role

Read the result

Higher = more time saved. Both groups benefit; the compound group benefits more.

Both groups share the sounds bij. What could explain the extra benefit?

What must a model explain?

• Units: preparing a morpheme saves more time than preparing its sounds alone.
• Order: in further experiments, a shared later morpheme, such as rol in bijrol, gave no preparation benefit.

What can be completed before the cue, and why must preparation start at the beginning?

WEAVER++: select, then encode

Concept → lemma + features → morphemes → sounds → speech

• A lemma supplies grammatical information; normally, only the selected lemma starts form encoding.
• Verification checks that a candidate belongs to the selected word. Activation affects how long selection takes.
• Rightward incrementality: build the sound plan from the beginning, pause, then resume when more is known.

For bijrol: prepare the morpheme bij and its syllable.
For bijbel: prepare the syllable only.

Dell: activation with feedback

Conceptual features ↔︎ lemmas ↔︎ morphemes ↔︎ sounds

• Activation spreads down and back up; shared sounds can support a word.
• At scheduled selection times, the most active eligible unit fills a slot.
• This helps explain mixed errors: rat for cat shares meaning and sounds.

Same units, different selection mechanisms

	Classical Dell	WEAVER++
Sounds feed back to word selection?	Yes	No
Selection	Most active eligible unit at a scheduled time	Activation plus verification; time varies
Main evidence explained	Speech-error patterns	Production-time patterns

Which mechanism explains why advance preparation changes response time?

Your Questions

Try it: prepare a grammatical form

Open the experiment · Choose Grammar

See GO and “Yesterday, I ___ home.” Say only went.

Condition	Example responses	Shared
Same grammatical context	went, ate, bought	Simple past; no common affix
Mixed grammatical contexts	went, eat, bought	Past, habitual present, present perfect

Was it easier to stay in one grammatical context?

In a shared past-tense block, you knew past tense before seeing the verb. The responses went, ate, bought share no common affix.

Did switching grammatical contexts make it harder to get ready?

Model question: can we prepare grammatical information before its word form?