Language History Questionnaire lhq3 An Enhanced Tool For Assessing Multilingual Experience
Language History Questionnaire lhq3 An Enhanced Tool For Assessing Multilingual Experience
Language History Questionnaire lhq3 An Enhanced Tool For Assessing Multilingual Experience
Cognition
enhanced tool for assessing
cambridge.org/bil multilingual experience
Ping Li, Fan Zhang, Anya Yu and Xiaowei Zhao
Review Article Pennsylvania State University and Emmanuel College
Introduction
Most researchers in the field of bilingualism or multilingualism need to use some form of self-
reported language history questionnaires, either abbreviated or extended. The purpose of the
questionnaire is to assess the linguistic background of bilinguals or second language learners so
that the researcher has a way of generating self-reported linguistic measures in multiple lan-
guages. The outcomes from such assessments are often used as independent variables to pre-
dict or to correlate with learners’ linguistic performances derived from behavioral or
neuroimaging experiments. Many language history questionnaires have been developed
(e.g., Marian, Blumenfeld & Kaushanskaya, 2007; Anderson, Mak, Chahi & Bialystok,
2018), including a standardized generic LHQ based on the most commonly asked questions
in previous published studies (Li, Sepanski & Zhao, 2006). LHQ and its successor (LHQ
2.0, Li, Zhang, Tsai & Puls, 2014) have now become widely used by scholars in bilingualism
and multilingualism research (e.g., more than 100 published studies have cited and used LHQ;
scholar.google.com indicates a total citation of over 300 times). Thanks to many researchers’
comments and suggestions in the past years, we have revised our instrument in its most recent
development (LHQ3), and this article provides a quick guide of its usage. The overall purpose
is consistent with the research community’s efforts to document and quantify bilingual and
multilingual experiences of the learner and the speaker (see the recent Anderson et al.,
2018 questionnaire).
of LHQ, or just certain items from the 22 questions. Participants’ (1) The “Download raw LHQ result” button will save partici-
responses to the questionnaire are automatically stored on the pants’ full raw responses in an Excel spreadsheet (in CSV
cloud server for 60 days, and can be retrieved as Excel spread- format);
sheets. LHQ 2.0 also provided researchers with data privacy (2) The “Download LHQ statistical data” saves the basic descrip-
through a password-protected account management interface. tive statistics for every item of the questionnaire across the
Through the management interface, investigators can retrieve participants in an Excel spreadsheet (also in CSV format).
and delete the data, and update user and experiment information. Specifically, the minimum, maximum, mean and standard
Finally, LHQ 2.0 also has multiple language functions so that deviations are automatically calculated for the questions
users could fill out the questionnaire in languages other than that involve data with numerical variables (e.g., age, self-rated
English (e.g., Chinese, Farsi, French, German, Spanish, proficiency, learning hours…), while the frequency distribu-
Turkish), although some of these versions did not have web-based tions are provided for the questions that involve data on a
interfaces (only PDF files for users to fill out on paper). nominal scale (e.g., gender, education, language).
LHQ 2.0 has been (and is still being) used by many researchers (3) The “Download LHQ aggregate scores” button will further
for their studies. Since 2014, we have received many inquiries, automatically calculate overall measurements of each partici-
comments, and requests for potential new functions to make pant’s language background and save the results in an Excel
the web interface more useful and powerful. To address the spreadsheet. The aggregated scoring function is important
needs of the researchers, we have developed LHQ3, which and we discuss the details below.
includes several improved new features while keeping all the
popular functions of LHQ 2.01.
Aggregated scores
New features of LHQ 3.0 Many researchers/users have requested that we develop an auto-
LHQ 3.0 includes several new features, including an enhanced matic scoring system to enable better and more user-friendly
account management interface, new data management interface, functions for LHQ (e.g., see comment in Anderson et al., 2018).
and new aggregated scoring functions. We discuss each of these Thus, in LHQ3, we introduced four aggregated scores to represent
below. participants’ overall proficiency, dominance, and immersion
levels of each language they have learned. It is our hope that
these scores will help the researcher to arrive promptly at a useful
Enhanced account management interface estimation/ classification of different types of multilingual
speakers.
Using LHQ 2.0, researchers can manage their projects through an
account management system interface (see Li et al., 2014).
Language proficiency
However, the interface was experiment-based, which means a
Multilingual speakers’ proficiency level in each language has been
researcher has to go through the signup process for every single
treated as an important dimension in many previous bilingual or
experiment and remember different IDs and passwords for differ-
multilingual studies (i.e., Bialystok, McBride-Chang & Luk, 2005;
ent experiments. This set-up could be an inconvenience to those
Chen, Zhou, Uchikoshi & Bunge, 2014). Taking this into consid-
researchers who had multiple questionnaires running, and, as a
eration, LHQ3 provides one overall aggregated scoring of profi-
result, researchers often asked the authors of LHQ for their for-
ciency based on the weighted sum of a participant’s self-rating
gotten IDs and passwords. To avoid this problem, we have
of his proficiency levels on different components of a language
designed a new user-based interface, which allows a researcher
(Question 11 of LHQ 3.0: Rate your current ability in terms of lis-
to manage all her projects under a single password-protected
tening, speaking, reading, and writing in each of the languages you
account (Figure 1). Through the interface the researcher can cre-
have studied or learned). So a participant’s overall proficiency
ate new questionnaires (with the option of choosing either Full,
score of his ith language can be written as:
Modular or Itemized LHQ), delete old ones, temporally disable
data collection of one questionnaire, download experiment 1
roaster, and access the data management page of the question- Proficiency i = vP
j={R,L,W,S} j i,j
(1)
7
naire (see next subsection). This enhanced system can give
researchers the flexibility in a trouble-free account management Here, {R, L,W, S} stands for Reading, Listening, Writing and
environment. Speaking components of a language. Pi,j stands for a participant’s
self-rated proficiency level to the jth component of his ith lan-
guage. Since it is rated on a 7-point Likert scale, we use a scaling
New data management interface
factor of 1/7 to normalize it into a range between zero and one
In addition to better account management, LHQ 3.0 includes an (with 1 indicating the native language-like proficiency level). ωj
all new data management interface (see Figure 2 for illustration), represents a weight assigned to the jth linguistic component.
which can be accessed by clicking the “Result” button of a project The default values of the weights for the four components are
on Figure 1. On this interface, participants’ responses to a ques- equally distributed (i.e., 25% each), but the investigator can desig-
tionnaire can be displayed for a quick check by the researcher. nate the weighting distribution for the four components on the
In addition, the researcher can download the responses to their data management interface (See Figure 3). The rationale is that
computers via three command buttons: different researchers may have different needs on which compo-
nent should be emphasized in their studies, therefore it is more
1
For the sake of brevity, the inherited features from LHQ 2.0, such as the sign-up pro- flexible for a researcher to use a weighted aggregated score in
cess, are not discussed here. Interested readers should refer to Li et at. (2014). given instances: for example, a study focusing on illiterate
bilinguals may only consider speaking and listening, with 50% of immersion into each language they have learned (Kuhl,
weighting to each of them (0% for reading and writing). Vice Stevenson, Corrigan, van den Bosch, Can & Richards, 2016).
versa, if a researcher is interested only in reading and writing in LHQ3 now introduces an overall aggregated scoring function of
the second language, then the reverse could be applied. immersion for each language that the participant knows, based
Appendix A provides a numerical example. on her Age, Age of Acquisition (AoA), and Years of Use of
the language (Question 5 of LHQ 3.0: Indicate your native lan-
Language immersion guage(s) and any other languages you have studied or learned,
Many researchers believe that immersion in a target language the age at which you started using each language in terms of lis-
environment might be the key to successful learning of a second tening, speaking, reading, and writing, and the total number of
language (Nikolov & Djigunović, 2006), and there is also empir- years you have spent using each language), using the following
ical evidence that immersion experience may attenuate L1 inter- equation:
ference to the L2 for late adult learners (Linck, Kroll &
Sunderman, 2009) and lead to corresponding brain changes in 1 Age − AOAi,j YoU i
the subcortical structures (Pliatsikas, DeLuca, Moschopoulou & Immersion i = v j +
2 j={R,L,W,S} Age Age
Saddy, 2017). It is therefore important for an investigator to
have a rough estimate of their multilingual participants’ history (2)
1 Pi,j 1 H i,j
Dominance i = v
j={R,L,W,S} j
+ (3)
2 7 2 K
Fig. 4. An output example of aggregated scores from a de-identified real participant (demographic information removed).
calculated aggregated scores clearly capture these participants’ dif- questions in LHQ3 generally take a serious adult participant
ference (see Appendix on how we calculate these scores). For 20–30 minutes to complete based on our experience (the Full ver-
example, participant yrgi4 (highlighted, 3rd row in Fig. 4) is a sion). Of course, as indicated above, researchers may modularize
Chinese speaker born in Hong Kong who learned 3 languages/ or itemize the LHQ with fewer questions, which would require
dialects (Cantonese L1; English L2; and Mandarin Chinese as less time to complete. LHQ has been successfully administrated
L3). The participant is more fluent in L2 and uses L2 more to both adults and adolescents. It was also adapted for other
often, and this is clearly captured by his higher L2 scores on pro- age groups including children (in translated or adapted versions
ficiency (column G: 1), immersion (column K: 0.72), dominance through parents’ reports; see Kambanaros & Grohmann, 2013;
score (column O: 0.59) and dominance ratio (column S: 1.38). Karpava & Grohmann, 2012; Yeong & Liow, 2012). Assistance
Participant jkdk8 (highlighted, 4th row in Fig. 4) is a Chinese may be required with participants who are not familiar with web-
speaker living in Beijing who learned 4 languages (Chinese L1; based interface.
English L2; Japanese L3 and French L4). He is obviously more LHQ 3.0 can be found at http://blclab.org/lhq3. As in previous
dominant on L1, as shown by his descending dominance ratios versions, LHQ3 has multilingual versions and can be completed
corresponding to the four languages (see columns R to U on in other languages. Currently, twelve language versions of LHQ3
the 4th row). These aggregated scores match with the partici- (English, simplified Chinese, traditional Chinese, French, German,
pants’ reports to individual questions from the LHQ. More Korean, Italian, Japanese, Portuguese, Slovenian, Spanish, and
importantly, their scores on an objective measurement of their Turkish) are available for use (more on the way). Researchers can
L2 (English) ability (Peabody Picture Vocabulary Test: follow four simple steps in deploying it for their study.
PPVT-4; Dunn & Dunn, 2007) are found to be significantly
correlated with their L2 proficiency score (r = .68, p =.001), L2 Step 1: The investigator completes the sign-up process (by clicking the
dominance score (r = .60, p =.007), and L2 dominance ratio to “Sign in/Sign up LHQ 3.0”) and logs on to the account manage interface
L1 (r = .71, p =.001). (Figure 1). Here, based on the investigator’s needs, the experimenter can
Finally, the use of the on-line web interface also makes LHQ3 create a full, modular, or itemized questionnaire with a unique
more reliable, since it can reduce potential transcription errors Experiment ID (with an individualized URL). The investigator will also
receive an email that contains the participant ID number roster in CSV
from manually recording and transcribing the data. We have
format. This is similar to the process in using LHQ 2.0.
also made further efforts to improve the reliability of LHQ 3.0 Step 2: The investigator can send this questionnaire URL and partici-
and to reduce errors from manual entry. For example, we imple- pant ID number (from the roster) to the participants from a study and ask
mented a synchronization function to ensure that a participant’s them to complete the questionnaire. The participants complete the LHQ
selection of the languages is consistent across multiple questions online through the individualized URL, and data (along with participant
in the data entry process. This function also saves participants’ numbers) are automatically stored and cumulatively saved. This is similar
time in filling the LHQ3 since their selection of languages in to LHQ 2.0.
the early questions can be automatically displayed later on with Step 3: The investigator accesses the account management page (by
questions that ask about languages. clicking the “Sign in/Sign up LHQ3”) to manage the questionnaires.
The investigator can delete a questionnaire and also temporally disable
or enable participants’ access to it. The account management system is
Practical steps on using LHQ3 updated and consolidated (as discussed earlier).
Step 4: At any time point, the investigator can reach the data manage-
LHQ3 is an instrument based on self-reports, and thus partici- ment interface (Figure 2) by clicking the “Result” button of a question-
pants need to understand the questions to complete it. The naire. Participants’ responses, basic descriptive statistics, and aggregated
scores will then be downloadable. The data management system is Jończyk R, Boutonnet B, Musiał K, Hoemann K and Thierry G (2016)
updated and made more flexible (as discussed earlier). The bilingual brain turns a blind eye to negative statements in the
second language. Cognitive, Affective, & Behavioral Neuroscience 16(3),
Finally, it should be noted that proficiency, immersion, and 527–540.
dominance are complex constructs in multilingualism research Kambanaros M and Grohmann KK (2013) Profiling (specific) language
impairment in bilingual children: Preliminary evidence from Cyprus.
and are often correlated with each other (see discussions in Luk
Solutions in the Assessment of Bilinguals 146–174.
& Bialystok, 2013; Treffers-Daller & Silva-Corvalán, 2015). The
Karpava S and Grohmann K (2012) Aspectual distinctions in bilingual
aggregated functions can provide the researchers with quick quan- Russian-Cypriot Greek children. Modern Greek Dialects and Linguistics
tifications of these constructs, but the researchers need to be mind- Theory 5(1), 124–140.
ful of their correlations, and need to evaluate a participant’s Kuhl PK, Stevenson J, Corrigan NM, van den Bosch JJF, Can DD and
language history in connection with other objective criteria or Richards T (2016) Neuroimaging of the bilingual brain: Structural brain
instruments for validity and reliability, as we discussed above, for correlates of listening and speaking in a second language. Brain and
a complete picture of an individual’s language background profile. Language 162, 1–9. http://doi.org/10.1016/j.bandl.2016. 07.004
Li P, Hsu C, Schloss B and Clariana R (2018) Acquisition of STEM concepts
Acknowledgements. Preparation of this article was made possible by a grant through expository text comprehension: An fMRI study. Poster presented at
from the National Science Foundation (BCS#1533625) to Ping Li. Partial the 24th Annual Meeting of the Organization for Human Brain Mapping,
support was also provided by the Shenzhen Peacock Team Plan Singapore, June, 2018.
(KQTD2015033016104926) and Guangdong Pearl River Talents Plan Li P, Sepanski S and Zhao X (2006) Language history questionnaires: A web-
Innovative and Entrepreneurial Team grant (2016ZT06S220). Fan Zhang based interface for bilingual research. Behavior Research Methods 38(2),
was supported by a Graduate Research Grant from Penn State University, 202–210.
and he is now with Interactions. LLC., Boston. We thank Rose Yuratovac, Li P, Zhang F, Tsai E and Puls B (2014) Language History Questionnaire
Zixin Tang, Chih-Ting Chang, Benjamin Schloss, Marissa Scotto and many (LHQ 2.0): A new dynamic web-based research tool. Bilingualism:
other members of BLC lab (http://blclab.org/) for helping test the online inter- Language and Cognition 17(3), 673–680. doi:10.1017/S1366728913000606
face. We are also grateful to many LHQ users for their comments and sugges- Linck JA, Kroll JF and Sunderman G (2009) Losing access to the native
tions that made the LHQ3 possible. language while immersed in a second language: Evidence for the role of
inhibition in second-language learning. Psychological Science 20(12),
1507–1515.
Luk G and Bialystok E (2013) Bilingualism is not a categorical variable:
References Interaction between language proficiency and usage. Journal of Cognitive
Anderson JA, Mak L, Chahi AK and Bialystok E (2018) The language and Psychology 25, 605–621.
social background questionnaire: Assessing degree of bilingualism in a Marian V, Blumenfeld HK and Kaushanskaya M (2007) The Language
diverse population. Behavior Research Methods 50, 250–263. Experience and Proficiency Questionnaire (LEAP-Q): Assessing language
Bialystok E, McBride-Chang C and Luk G (2005) Bilingualism, language pro- profiles in bilinguals and multilinguals. Journal of Speech, Language, and
ficiency, and learning to read in two writing systems. Journal of Educational Hearing Research 50(4), 940–967.
Psychology 97, 580–590. Malt B, Li P, Pavlenko A, Zhu H and Ameel E (2015) Bidirectional lexical
Bidelman GM, Gandour JT and Krishnan A (2011) Cross-domain effects interaction in late immersed Mandarin-English bilinguals. Journal of
of music and language experience on the representation of pitch in the Memory and Language 82, 86–104.
human auditory brainstem. Journal of Cognitive Neuroscience 23(2), McLeod S and Verdon S (2017) Tutorial: Speech assessment for multilin-
425–434. gual children who do not speak the same language (s) as the speech-
Bidelman GM, Hutka S and Moreno S (2013) Tone language speakers and language pathologist. American journal of Speech-Language-Pathology
musicians share enhanced perceptual and cognitive abilities for musical 26(3), 691–708.
pitch: evidence for bidirectionality between the domains of language and Nikolov M and Djigunović JM (2006) Recent research on age, second lan-
music. PloS One 8(4), e60676. guage acquisition, and early foreign language learning. Annual Review of
Calvo N, García AM, Manoiloff L and Ibáñez A (2016) Bilingualism and Applied Linguistics 26, 234–260.
cognitive reserve: a critical overview and a plea for methodological innova- Pliatsikas C, DeLuca V, Moschopoulou E and Saddy J (2017) Immersive
tions. Frontiers in Aging Neuroscience 7, 249. bilingualism reshapes the core of the brain. Brain Structure & Function
Carlson MT, Goldrick M, Blasingame M and Fink A (2016) Navigating con- 222, 1785–1795.
flicting phonotactic constraints in bilingual speech perception. Bilingualism: Treffers-Daller J and Silva-Corvalán C (eds.) (2015) Language dominance in
Language and Cognition 19(5), 939–954. bilinguals: Issues of measurement and operationalization. Cambridge:
Chandrasekaran B, Krishnan A and Gandour JT (2009) Relative influence of Cambridge University Press.
musical and linguistic experience on early cortical processing of pitch con- Yang J, Gates KM, Molenaar P and Li P (2015) Neural changes underlying
tours. Brain and Language 108(1), 1–9. successful second language word learning: An fMRI study. Journal of
Chen SH, Zhou Q, Uchikoshi Y and Bunge SA (2014) Variations on the Neurolinguistics 33, 29–49.
bilingual advantage? Links of Chinese and English proficiency to Chinese Yeong SH and Liow SJR (2012) Development of phonological awareness in
American children’s self-regulation. Frontiers in Psychology 5, 1069 English–Mandarin bilinguals: A comparison of English–L1 and
Dong Y and Zhong F (2017) Interpreting experience enhances early atten- Mandarin–L1 kindergarten children. Journal of Experimental Child
tional processing, conflict monitoring and interference suppression along Psychology 112(2), 111–126.
the time course of processing. Neuropsychologia 95, 193–203.
Dunn DM and Dunn LM (2007) Peabody Picture Vocabulary Test (PPVT),
4th Edition, Manual. Minneapolis: NCS Pearson, Inc.
Grant A and Li P (2019) Proficiency affects intra- and inter-regional BOLD
Appendix A: Example of LHQ3 Data and Analyses
variability in second language processing. (in press, Language, Cognition,
and Neuroscience) The following example is provided to help the readers understand how the
Hartanto A and Yang H (2016) Disparate bilingual experiences modulate aggregated functions work. Imagine a Chinese-English bilingual at the age
task-switching advantages: A diffusion-model analysis of the effects of of 25 and with English as his L2, his answers to many questions of LHQ
interactional context on switch costs. Cognition 150, 10–19. are listed below.
L1 (Chinese) L2 (English)
Assume that a researcher treats different linguistic aspects (listening, speaking, And his immersion score on English (L2) would be 0.52
reading and writing) equally and assign each aspect a weight of .25 (i.e., 25%).
The participant’s overall proficiency score on Chinese (L1) would be 1. Immersion L2
1 1 1 Age−AOAi,j YoU i
Proficiency L1 = vj Pi,j = (0.25∗ 7 + 0.25∗ 7 + 0.25∗ 7 + 0.25∗ 7) = 1 = vj +
7 j={R,L,W,S} 7 2 j={R,L,W,S} Age Age
1 25−9 25−9 25−9 25−9 10
= 0.25∗ +0.25∗ +0.25∗ +0.25∗ +
And his proficiency score on English (L2) would be 0.75 2 25 25 25 25 25
1 64 10
1 1 = + = 0.52
Proficiency L2 = vj Pi,j = (0.25∗ 6 + 0.25∗ 6 + 0.25∗ 5 + 0.25∗ 4) = 0.75 2 100 25
7 j={R,L,W,S} 7
When his daily use of the languages is considered, we can also calculate his
Please note that a researcher could give different weights to different lin- dominance score for the two languages.
guistic aspects for the research purpose. For example, if a researcher only
focuses on bilingual participants’ listening and speaking components of the
1 Pi,j 1 H i,j
two languages, he could set the weight to be 0.5 each for listening and speak- Dominance L1 = vj +
j={R,L,W,S} 2 7 2 K
ing, but 0 for the other two (reading and writing). In this case, the participant’s
overall proficiency score on Chinese (L1) would still be 1. 1 7 1 8 1 7 1 8
= 0.25 + + 0.25 +
2 7 2 16 2 7 2 16
1
Proficiency L1 = (0.5∗ 7 + 0.5∗ 7 + 0∗ 7 + 0∗ 7) = 1 1 7 1 4
7 + 0.25 +
2 7 2 16
1 7 1 2
+ 0.25 + = 0.671875
And his proficiency score on English (L2) would be 0.86, larger than last 2 7 2 16
calculation when reading and writing were considered. This makes sense
since he is not good at reading and writing in L2.
1 1 Pi,j 1 H i,j
Proficiency L2 = (0.5∗ 6 + 0.5∗ 6 + 0∗ 5 + 0∗ 4) = 0.86 Dominance L2 = v +
7 j={R,L,W,S} j
2 7 2 K
1 6 1 4 1 6 1 4
= 0.25 + + 0.25 +
We can also calculate his immersion scores. For his L1 (Chinese), it would 2 7 2 16 2 7 2 16
be 0.93
1 5 1 2
+ 0.25 +
2 7 2 16
Immersion L1 1 4 1 1
+ 0.25 + = 0.4609
1 Age − AOAi,j YoU i 2 7 2 16
= vj +
2 j={R,L,W,S} Age Age
Therefore the Ratio of Dominance for L1 will be 1 and the Ratio of
1 25 − 1 25 − 2 25 − 3 25 − 4 24 Dominance for L2 will be
= 0.25∗ + 0.25∗ + 0.25∗ + 0.25∗ +
2 25 25 25 25 25
1 90 24 Dominance L2 0.4609
= + = 0.93 RatioDominanceL2 = = = 0.686
2 100 25 DominanceL1 0.671875