To what extent does acquiring a second language also mean acquiring a second language behaviour? This study tests the hypothesis that fluent speakers of more than one language display different cognitive patterns when performing word association tasks in different languages.
Seven Mandarin-English bilingual participants performed a simple word association task in both English and Mandarin in which they rapidly spoke words that they associated with each of a series of terms describing rare or imaginary animals.
As a control, the same task was performed in English by seven monolingual English native speakers.
Responses were then divided into three separate pairs according to language and speakers’ background: (i) Englishes (monolinguals’ and bilinguals’ responses in English), (ii) L1s (responses given by participants in their first language), (iii) Bilinguals (bilingual participants’ responses in both languages).
Similarities were quantified using Hellinger Affinity index. Interesting data emerged, showing a significantly higher affinity score within the Englishes compared to the L1s.
This suggests that bilingual speakers performed different kinds of associations to the same animal according to what language they were asked to speak at the time: their word associations when speaking English were more similar to those performed by native English speakers if compared to the word associations they performed when speaking Mandarin Chinese. As shown in the following examples, the affinity score for Englishes is higher than the affinity score for L1s: crocodile (Englishes: 0.72, L1s: 0.528, Bilinguals: 0.771), panda (Englishes: 0.734, L1s: 0.679, Bilinguals: 0.704), lion (Englishes: 0.732, L1s: 0.478, Bilinguals: 0.886), bear (Englishes: 0.678, L1s: 0.405, Bilinguals: 0.623), dragon (Englishes: 0.578, L1s: 0.263, Bilinguals: 0.544).
1. Sara Bologna, Karen Sullivan,
The University of Queensland
sara.bologna@uq.net.au, ksull@uq.edu.auu
BILINGUALISM AND BICULTURALISM: A PILOT
STUDY ON WORD ASSOCIATION TASKS
PERFORMED BY BILINGUALS
2. The background - overview
Do speakers of more than one language show different behaviours
according to which language they are speaking at a given time?
● Keysar et al. (2012). The foreign-language effect: thinking in a
foreign tongue reduces decision biases.
● Costa et al. (2014). Your morals depend on language.
● Yoon (2004). Category norms as a function of culture and age:
comparisons of item responses to 105 categories by American
and Chinese adults’
3. The background - Keysar et al.
● Keysar et al. (2012). The foreign-language effect: thinking in a
foreign tongue reduces decision biases.
Four different studies show how decision biases are reduced when
a choice is presented in a foreign language, due to the
disappearance of the framing effect.
4. The background - Costa et al.
● Costa et al. (2014). Your morals depend on language.
Presented evidence of a difference in morality judgements when
performed in one’s second language rather than their first
language.
5. The background - Yoon et al.
● Yoon et al. (2004). Category norms as a function of culture and
age: comparisons of item responses to 105 categories by
American and Chinese adults
○ Standard Category Fluency Tasks to compare cross-
cultural responses between Mandarin and American
English speakers
○ Participants’ cultural background had a significant impact
on their responses
6. The question
To what extent does acquiring a second language also mean
acquiring a second language behaviour?
7. The hypothesis
Fluent speakers of more than one language display different
cognitive patterns when performing word association tasks in
different languages.
8. The study - overview
● Method
○ Participants performed a simple word association task in
both English and Mandarin
○ Control: the same task was performed in English by
monolingual English native speakers
○ Responses were divided into three groups: Englishes, L1s,
Bilinguals
○ Hellinger index used to quantify similarities
9. The study - participants
● Six bilingual adults (Mandarin and English)
● Similar education background
● Lived in China for first 10+ years of their life
● Lived in AUstralia for 4+ years
● Completed 1+ years of tertiary studies in Australia
● Use both languages on a daily basis
● Visit China regularly
10. The study - control group
● Six monolingual adults (Australian English)
● Born and raised in Australia by English speaking families
● No recent long stay in non-English speaking countries
● No fluency in any other language
● Never exposed to other languages during school years
11. The study - data collection
● The task: spontaneous word associations
○ Cues: 23 words describing animals
○ 30 seconds to perform as many associations as possible
○ Bilinguals performed task in both languages
■ Long break and distracting activities in between
○ Data grouped in semantic categories and translated
12. The study - data analysis
● Responses to cues divided into 3 groups: EE, L1s, Bilinguals
● Each cue analysed individually to identify response probability
● Similarity between groups calculated: HAI
13. The study - results
● For rare and imaginary animals: significantly higher affinity
score within the EE compared to the L1s
○ Crocodile: EE: 0.72, L1s: 0.528, Bilinguals: 0.771
○ Panda: EE: 0.734, L1s: 0.679, Bilinguals: 0.704
○ Lion: EE: 0.732, L1s: 0.478, Bilinguals: 0.886
○ Bear: EE: 0.678, L1s: 0.405, Bilinguals: 0.623
○ Dragon EE: 0.578, L1s: 0.263, Bilinguals: 0.544
14. Conclusions - what is interesting
● For rare and imaginary animals: significantly higher affinity
scores within the EE compared to the L1s
→ Bilingual speakers performed different kinds of associations to
the same animal according to what language they were asked to
speak at the time
→ This is only true for rare and imaginary animals: maybe because
we learn about this animals through language rather than through
personal experience?
15. Conclusions - scope for further research
● Results are only just significant: study should be replicated on a
larger sample
● Further research might look into different aspects, e.g.
○ How do different cues give rise to different similarities?
○ Would similar results arise from different language
combinations?
○ Would results still be significant across different age
groups?
16. Sara Bologna, Karen Sullivan,
The University of Queensland
sara.bologna@uq.net.au, k.sull@uq.edu.auu
Thank You.
Questions?
Editor's Notes
Recent research in cognitive psychology and psycholinguistics has investigated whether speakers of more than one language show different behaviours according to which language they are speaking at a given time. Interesting results have been published by Keysar et al. (2012) and Costa et al. (2014) with regards to the so called Foreign Language Effect, which supposedly makes reasoning in a foreign language more rational due to the speaker’s lower emotional involvement with their second language as opposed to their first language. Yoon’s study showed evidence of how the participants’ cultural background can impact their responses in standard category fluency tasks.
Keysar conducted four different studies showing how decision biases are reduced when a choice is presented in a foreign language, due to the disappearance of the framing effect. Whereas people were risk averse for gains and risk seeking for losses when choices were presented in their native tongue, they were not influenced by this framing manipulation in a foreign language. The authors propose that these effects arose because a foreign language provides greater cognitive and emotional distance than a native tongue does.
Costa et al. (2014, Experiment 1) presented participants in four different countries with the Footbridge problem in various native and foreign languages. They found that those using a foreign language gave significantly more utilitarian responses than those using a native one (with the only exception being East Asians, which is consistent with previous findings; Gold, Colman & Pulford, 2014). In Experiment 2, the participants were given both Switch and Footbridge dilemmas in a native or foreign language. Crucially, there were two samples: one had English as a native language and Spanish as a foreign one, and the other had the inverse pattern of languages. The results replicated the effect of language on responses to Footbridge, but found no effect of language for Switch.
Yoon (2004) used standard category fluency tasks to perform a cross-cultural comparison between Mandarin and American English speakers, finding that the participants’ cultural background had a significant impact on their responses. However, Yoon’s study did not look at bilingual subjects and this raises the question of whether such cultural differences would be reflected in the answers of bilingual speakers and what would be the influence of the language they are using to perform the task on their responses.
Standard category fluency tasks: used in cognitive psychology and psychiatry to test for fluency through indices of vocabulary size and lexical access speed. In these standardised tests participants are given a broad semantic category as a cue (e.g. food) and are asked to produce as many words (or responses) as possible within that category (e.g. fruit, pizza, chocolate) in a time frame of sixty seconds.
The studies just mentioned, and many others, have been conducted on large numbers of late learners of a second language who did not acquire native-like proficiency, and only a very limited number of similar studies conducted on fluent bilinguals are currently available. Most literature addressing the bilingual mind comes from cognitive psychology and is interested in a variety of topics such as advantages and disadvantages in the cognitive development of bilingual children, the effects of bilingualism on cognition during adulthood or if and how bilingualism is correlated to the onset of mental disorders later in life, failing to explore other aspects.
In light of previous literature analysed on the topic, this study tests the hypothesis that fluent speakers of more than one language display different cognitive patterns when performing word association tasks in different languages. Specifically, we expect bilingual speakers of Mandarin Chinese and English to associate different concepts to the same cue, according to which language the cue is provided in. Cues are words referring to animals, specifically rare of imaginary animals. This is because we learn about these animals through language rather than through personal experience.
Mandarin-English bilingual participants performed a simple word association task in both English and Mandarin in which they rapidly spoke words that they associated with each of a series of terms describing animals.
As a control, the same task was performed in English by monolingual English native speakers.
Responses were then divided into three separate pairs according to language and speakers’ background: (i) Englishes (monolinguals’ and bilinguals’ responses in English), (ii) L1s (responses given by participants in their first language), (iii) Bilinguals (bilingual participants’ responses in both languages).
similarities were quantified using Hellinger Affinity index.
Word association task: They rapidly spoke words that they associated with each of a series of terms describing rare or imaginary animals.
Hellinger Affinity index: used to quantify the similarity between two probability distributions.
To ensure an appropriate language fluency level in the Chinese subjects, specific criteria were enforced during the recruiting process: all bilingual participants are native speakers of Mandarin Chinese who lived in China for at least the first ten years of their life, have lived in Australia for at least four years, have completed at least one year of tertiary studies in Australia, use both languages on a daily basis and visit China regularly.
Gender ratio: 4 F, 2M
All monolingual participants were born and raised in Australia by English speaking families and have never lived in a non English speaking country for more than 6 months since 2009. With the exception of short periods of time abroad, the occasional language course and some self-directed foreign language studies, none of the monolingual participants has reached complete fluency in a language other than English and none of them was exposed to any language other than English before high school.
Gender ratio: 3F, 3M
Because this study is not strictly interested in fluency, but in comparing spontaneous word associations between monolingual and multilingual individuals, two significant changes were brought to the Standard Category Fluency Task format:
Because words often don’t have a perfect correspondence from language to language but present differences in connotation and use (Mushangwe, 2014), the list of cues selected for this study consisted of 23 nouns referring to animals that are well known in both the English speaking and Chinese culture
the time frame was halved to 30 seconds per cue, on the assumption that the very first words uttered or written would be closer to an instinctive association rather than a more rational mental search for a list of terms. A reduced time frame also reduces the influence of chaining, whereby later associations are made based on previous associations, rather than based on the original cue
The monolingual English speakers only performed the task once in their native language, while the bilingual participants performed it in two sessions (one for English, one for Mandarin) with a fifteen to twenty minute interval between sessions. During the interval participants were engaged in distracting activities (small talk, sudoku puzzles and arithmetics) aimed at minimising the chances of them remembering and replicating the responses given in the previous session.
After data collection, the responses were cleaned and grouped into larger semantic categories. This was done to account for synonimity and to remove translation bias. A certified NAATI translator translated all the Mandarin responses into English and provided a detailed explanation of cultural references where needed.
Monolinguals and bilinguals responding in English (EE), participants and control group memebers responding in their native language (L1s), and bilinguals responding in both languages (bilinguals).
For each group a list of all responses to the cue from all participants were combined into a list. The number of occurrences of each unique response was counted. Responses occurring only once were removed. The remaining responses were converted to a fraction of the total number of responses given by the group. This can be treated as the probability that the response will be given for the cue in that group (For example, the bilingual Chinese speaking Mandarin group (CM) generated a total of 17 responses for the cue Swan, comprised of 8 unique words. Of the 17 responses, 3 were instances of the adjective beautiful, so beautiful is given the probability of 3/17)
For each unique response to the cue, the similarity between different speaker groups could then be calculated. The measure of similarity used was the Hellinger Affinity index, a value in the range [0,1] calculated as the square root of the product of the probability of a word being uttered in each group. The greater the difference in probability for a particular response, the smaller the Hellinger Affinity index number. For each response in the cue, the Hellinger Affinity index was calculated for each of the three pairs of groups. For each pair of groups, an overall similarity score for the cue is determined by summing the Hellinger Affinity indexes of all unique responses.
In regards to rare and imaginary animals, bilingual speakers performed different kinds of associations to the same animal according to what language they were asked to speak at the time: their word associations when responding in English were more similar to those performed by native English speakers if compared to the word associations they performed when speaking Mandarin Chinese. Results show how the affinity index for EE (both monolinguals and bilinguals speaking English) is higher than L1s (participants and control group members responding in their native language)
A more extensive investigation into how different cues give rise to different similarities between groups would also be a worthy avenue of research.