Notes on The Influence of Sentence Context Constraint on Cognate Effects in Lexical Decision
and Translation by Janet G. Van Hell.
1. Cross-language activation, cognate effect and linguistic context
In linguistic, two main theoretical views are taking place on how bilinguals activate and
process words in their two languages: the language selective activation and the language-
nonselective activation. The language selective activation theory think that bilingual are
activating words from the context appropriate language when language nonselective theory
think that bilingual are activating words from both languages.
Most of studies have shown that words are activated in parallel and so evidence for
language nonselective activation is predominant. However, most of those studies have been
done in isolation, without a linguistic context (no external help to discover the word/ picture).
The research in this paper assumed that bilingual are activating both language at the
same time, a language nonselective activation process. For that, Janet G. Van Hell used
cognate and non-cognate words. Cognate words are words that have the same common
etymological origin (same orthography, semantic and phonology), in this study same origin in
English and Dutch, that leads to a faster recognition and production times of cognate words
over non-cognate ones.
To expand the previous research done (out of a meaningful linguistic context), the author
would study the use of cognate and non-cognate words in different context: isolation, high
contextualized sentences, and low contextualized sentences.
Those kinds of studies have been done with monolingual speakers and the result show
that context information influences words processing. Words in a constraining context are
recognized faster than words in neutral context. Also in those studies on monolingual speakers,
in isolation, concreteness, ambiguity, or frequency are modulated by sentence context: concrete
words would be processed faster than abstract ones. This difference disappear in a highly
constraint sentence context. The same appear to be right with frequency words and semantic
ambiguity. (In isolation great difference, the difference is reduced largely in highly constraint
context). Sentence context is highly predictive of the outcome whatever the word but not all
of them are. In natural language situation; the context is only midly predictive.
The 3 experiments are studying “whether sentence context modulates cognate effect,
and the role of semantic constraint of the sentence context”. The three studies are targeting
congregate and non-congregate words in isolation, high constraint sentence, and low constraint
The participants in the study were fairly proficient in English, and their first language was
Dutch. They learn English starting at 10 to their entrance to university in which most of their
readings were in English. Lexicon in English was smaller than their native language lexicon.
They were all first-year student at the University.
2. Sentence context norming studies Two norming studies were performed to determine the degree of contextual constraint and
the plausibility of the sentence contexts to be used.
a. Sentence completion study
The first studies was done in order to qualify whether the sentence were high or low
constraint. 240 participants from the same population of the study were asked to write
down the first three reasonable completions that came in mind while present with a one
word omitted sentence.
b. Sentence plausibility ratings
In order to reduce the plausibility of the sentences proposed (more plausible= better
performance and so it may interfere with the high constraint of context). 240 from the
same population were given the sentences and the intended target word, and were
asked about the plausibility.
3. Experiment 1: lexical decision in the second language
A 2 (cognate vs. noncognate) by 3 (high constraint, low constraint or no context) factorial
design was used.
The participants are 60 fairly fluent bilingual randomly allocated to one of the three
context conditions. The mean of comprehension and production was fairly equal for the
60 words were used (previously studied with the norming groups above): 30 cognates
and 30 noncognates. Half of each category was concrete words and half was abstract.
Then they were separated in 3 differents categ