The Influence of Sentence Context Constraint on Cognate Effects in Lexical
Decision and Translation
Janet G. van Hell
- Main question in bilingual research: How bilinguals activate and process words in their two languages, and
how do they control their two language systems.
o Two theories:
- Language-selective activation view – bilinguals activate only contextually relevant word
candidates from the language.
- Language-nonselective activation – words from both languages are activated in response
to incoming information (nonselected information).
o More studies point toward the activation of words in bilingual memory as operating
in a language-nonselective way. Most studies give participants pictures and are asked to
name what is depicted.
o Study: Dijkstra, Van Jaarsveld & Ten Brinke (1998) Cognate Effect
- Dutch-English bilinguals
- Lexical tasks using a set of identical cognates (same word and meaning in both
languages), a set of identical interlexical homographs (words with an identical spelling,
but different meaning in the two languages) and matched controls e.g. noncognates.
- Results: lexical decision times on the cognates shorter than those on the matched
controls. The difference in decision time between the interlexical homographs and their
controls varied across experiments; sometimes faster, slower than or as fast as. Depended
on task demands and stimulus list composition.
o Study replicated one year later (1999) by Dijkstra, Grainger, & Van Heuven
- Multilinguals – Dutch (Native), English (L2), French (L3)
- Similar cognate effects have been observed in a wide range of tasks
- Obtained a facilitation effect on L1 words (e.g., the Dutch word ‘appel’) that were
cognates with their L2 translation (English, ‘apple’) or with their L3 translation (e.g., the
Dutch word ‘meubel’, meaning ‘meuble’ in French, and ‘piece of furniture’ in English.
- This was obtained only with speakers who were relatively fluent in their L3
o The advantage of cognates over noncognates when presented in isolation
- a language-nonselective activation process in which word candidates from both
languages are activated in parallel
- The presentation of a word in one language co-activates words in the other language,
most likely words that are highly similar in orthography, semantics, and phonology.
- Cognates are the same in both languages leading to faster recognition and reproduction
times for cognates.
- This cognate effect fits in with recent models of isolated-word processing in bilingual
memory: The inhibitory control model and the Bilingual Interactive Activation BIA(+)
o Aforementioned studies all studied the processing of single words presented in isolation (out of
meaningful linguistic context). Not a common task o As well, the linguistic context surrounding the target word may influence the lexical activation
process of the word. E.g. the sentence context containing the target word may influence the
degree of co-activation of related words in bilingual memory
o How words are recognized in natural sentence contexts = a critical test of the validity of the
bilingual isolated word recognition and reproduction models.
o Studies investigating the processing of target words presented in a linguistic context found that
contextual information influences word processing. This is done with monolinguals as research
with bilinguals in scarce. E.g. words (e.g., ‘apple’) that are preceded by a semantically
constraining context (like ‘She took a bite of the fresh green ….’) are recognized faster than words
preceded by a neutral sentence context (like ‘The final word in this sentence is ….’).
o Studies with monolingual speakers also showed that effects of word type, like concreteness,
frequency or ambiguity, are modulated by sentence context. Performance on concrete words
better. Highly predictable or frequent words, best.
o The proficiency level of the bilinguals participating in the study: all unbalanced bilinguals with
Dutch as their native language and English as their second, and all were first-year students of the
University of Amsterdam, the Netherlands. Starting learning English at around age 10, all attended
English classes at secondary school for about 3-4 hours a week, until university. The participants
were fairly proficient in their second language.
Sentence context norming studies
1. Sentence completion study
• 240 participants
• Presented with sentences with one omitted word and asked to write down the first 3 reasonable
• 2 constraint conditions: high constraint contexts and low constraint contexts
• Half of the sentence contexts, for both constraint conditions, the omitted target was the final word
of the sentence; other half was located somewhere in the middle of the sentence.
• 480 sentence fragments presented in 8 different booklets
• All sentences (5 per page) in a book were in the same language (either Dutch or English) and were
at the same level of constraint (high or low).
• The production probability of each completed word was calculated across participants.
2. Sentence plausibility ratings
• Study assessing the plausibility of the sentence contexts
• Performance on sentences describing plausible events is better than on sentences describing less
• Difference in the plausibility of sentence contexts could possibly interfere with the influence of
sentence constraint on the upcoming words.
• New group of 240 participants
• They were presented with the complete sentences and were asked to assess the plausibility.
• It was stressed that the plausibility of the situation as described