09 Jul Psycholinguistic Experiments: Design, potential problems and addressing them
Psycholinguistics is one of the oldest experimental fields of linguistics. However, along with the developments in the field of technology, the research trends in this area have also gone through various changes.
While throwing light at previous trends, this survey paper tries to cover the current trends to design a psycholinguistic experiment, possible problems and the solutions to address them. A sample experiment is provided in the appendix section.
To avoid unnecessary expansion of subject matter, this paper deals with the experiments related to purely linguistic aspects such as word recognition and sentence processing.
Field of psycholinguistics is concerned about the factors that enable human beings to acquire, use, comprehend and produce language. The trends in psycholinguistics have gone through significant changes recently, as earlier the branch was mainly based on theoretical findings (Caramazza & Mccloskey, 1980), (O’Connell & Kowal, 2011). However, recent advancements in the fields of biology, neuroscience and information science have given birth to yet another aspect of psycholinguistics, which is grounded much more on experimental elements (Goodwin & Hein, 1982), (Kess, 1991).
The field developed initially as an exploratory discipline which tried to figure out the linguistic problems caused due to some accidental physical damages (later on monitored experiments), and afterwards, the focus got shifted to the exploration of language as a whole.
- Subfields which need experimental insights
We learn a language as a child in a speech community. Its such a part of life which goes completely unnoticed due to its inevitability. Most of the time the process of language acquisition is taken as for granted as there is no direct effort involved. However, this part of life is very crucial, and any impediments at this stage of language development affect the verbal behaviour throughout the lifetime. Other than this, there are special conditions which may give birth to many disorders in the early stage of language learning, and if they go unnoticed, they may affect the language system very severely.
Some disorders occur due to an accident after the language acquisition is complete. These may affect the verbal behaviour entirely or partially along with the other physical activities. All these disorders can impair the verbal behaviour considerably. However, early information and timely treatment may help reduce the impairment level to the lowest.
This way the principal subfields of psycholinguistics, which need experimental insights can be divided into three genres, language acquisition, language comprehension and language production. Further, a division in the areas of psycholinguistics can be made based on the involvement of intrinsic and extrinsic factors as:
- Linguistics-based areas
- Cognition-based areas
Linguistics related areas are primarily concerned with the language and its processing only. On the other hand, cognition related areas have a broader scope and include non-linguistic interests such as human and artificial intelligence, and human behaviour as well. In the interdisciplinary studies involving both of the fields, language processing is analysed diachronically along with the overall human behaviour by analysing the behavioural changes and responsible neural activities. In such cases, cognitive studies help by establishing the relation between the resultant behavioural changes during language processing and the related neuronal activities.
- Psycholinguistic experiments in a nutshell
Based on the requirements of a study, a psycholinguistic experiment can be straightforward and manual or very complicated along with the involvement of cutting edge technology and machinery these days. However, a core methodology is followed in every experiment. Each psycholinguistic experiment involves a tester or a scientist, who tests its hypothesis on multiple subjects through some linguistic data (which may be in various forms), records the results in a specific way and try to get some meaningful insights by them to get a conclusion (Walters & Wolf, 1986). These experiments help understand many unnoticed aspects of language, which in result may potentially assist in treating various disorders.
- Various experimental setups
Psycholinguistic setups can be utterly manual without the involvement of any machine and can have a significant role of machines in them. These are some of the experimental setups used for psycholinguistic experiments these days:
4.1 Baby lab
In such labs, infants’ social, cognitive, and linguistic development in the first three years of their life is investigated by the experts.
4.2 Behavioral lab
Behavioural labs are used to run behavioural experiments in which speech or manual responses to speech are recognised by a fully manual task. With the evolution of technology, now a day soundproof cabins, equipped with high-fidelity headphones, microphones, and pushbutton devices are used to yield accurate results from such experiments.
4.3 EEG lab
In EEG labs brain potentials are recorded while a subject is carrying out some given task (e.g., reading or listening to text, evaluating words or texts on some dimension, etc.). In such experiments, first of all, an electrode cap is fitted over subject’s head, and then measurements are done as it does its tasks. Sometimes the process is also used for Cognitive Neuroimaging which helps understand cognitive processes involved in carrying out a specific task.
4.4 Eyetracking lab
High-end eye tracking equipment is used in such labs now a day. Usually, a total of three eye trackers are used in such experiments out of which two are head mounted, and the other one can be mounted based on the specific requirements of the study. These trackers monitor eye movements while the assigned task is carried out by the subject.
4.5 fMRI lab
The fMRI labs are equipped with a fMRI scanner, which is basically a huge fancy magnet with a bed inside, where a subject can lay down as it carries out its task. In this machine, the brain activity is measured by Functional Magnetic Resonance Imaging technology as the changes in the blood flow of the subject take place. These tastes rely on the fact that when a specific area of the brain is involved in an activity, the blood flow increases there.
4.6 Gesture lab
Such labs are designed to conduct experiments which study gestures and sign language. These are equipped with audio recorders and multiple synchronised video cameras allowing the same scene to be recorded from many angles, which can later be used to diagnose various facial expressions and gestures.
4.7 MEG lab
MEG labs are equipped with a MEG machine which looks like a giant professional hairdryer and is positioned over the subject’s head while it is asked to carry out some task. Most of the MEG experiments end with a visit to the fMRI lab, where an anatomical picture of subject’s brain is taken. In the end, the findings are prepared from the magnetic activity and blood circulation reports of the brain to locate the area responsible for a particular type of brain activity.
4.8 Virtual reality lab
Virtual reality is a technology-enabled, machine generated and interactive experience that takes place in a simulated environment. Labs with such technology are still in the process where state of the art immersive Virtual Reality (VR) equipment are used for various purposes.
- Types of conducted experiments and their design
Majority of experiments conducted to date in psycholinguistics are focused on word recognition and Sentence processing. These experiments give insights into how we recognise words and how do we understand a sentence. Based on the two different forms of language representation, the experiments are done both for written and spoken form.
5.1 Word recognition
This can be divided further into two types based on the different mediums of representation of a language:
5.1.1 Visual word recognition
Visual word recognition is related to the recognition of arbitrary symbols, and the findings in such experiments are based on eyeball movement, facial expressions, and activities in various brain areas. The prime purpose of such experiments is to explore the mental semantic associations with the real world visual representation and their properties. Subjects in this type of experiments are given various word data sets, and the findings are drawn by one or combining more than one data sources listed above.
5.1.2 Spoken word recognition
Experiments based on spoken word recognition explore the auditory possibilities based on the nature of sound patterns and the way the human brain interacts with them to find the significant association with arbitrary sound units.
5.2 Sentence processing
This also can be divided into two types based on the different kinds of perception:
5.2.1 Visual sentences processing
The experiments to find the process involved in visual sentence processing are much similar to the visual word processing. The results are based on eyeball movement, facial expressions and the areas of the brain which gets activated during the whole process.
When a subject tries to comprehend a sentence in a language, there are many possibilities, which may result in uncertainties such that how the referred objects are connected to each other. This happens because when our eyeballs move in the direction of the script (left to right or right to left) the information required to establish correct dependencies between the lexical items is not available by the moment.
5.2.2 Spoken sentence processing-
A sentence when uttered is a string of arbitrary sounds which yields meaning in specific combinations. Interestingly the meaningful sounds are not discrete/separate but a part of the whole stream of sounds along with all other sounds present in the environment. Such things can be problematic in many cases and may yield exciting findings to explore the nature of our language faculty in many cases.
- Potential problems and the way to address them
A researcher can face many problems in a psycholinguistic experiment based on the different factors involved. While conducting psycholinguistic research, we must take into account the linguistic characteristics of the materials and other characteristics that might exert an influence on the conclusion. Some of the potential problems faced by the researchers are:
- Extraneous variables:
These are the causal variables in which a researcher is not interested due to their less or no importance, however, these influence the dependent variables and impair a valid causal conclusion. Therefore, based on the purpose and type of study, extra care should be taken of such variables. Some of such variables are:
- Age group
- Voice quality
- Socio-economical background
Other than these variables, the demographic variables of less importance and detailed sociolinguistic profiling of subjects may help in getting good results (Dietrich &Graumann, 1989).
- Word frequency and length
Different word frequencies and length in an experiment or two different experiment sets can alter the findings a lot, that is why one should keep this in mind while designing data/questionnaire for an experiment. The best option is to use the same data sets on different subjects, or a researcher should take special care of such factors while designing datasets.
- The transitional probability of n-grams
Not all sets of words have the same probability of utterance as other words. Therefore, it should be checked in the chosen data that most of the n-grams or at least bigrams, used for forming a hypothesis should have more or less similar transitional probability. In case of different transitional probabilities, the findings will suffer adversely which may affect the hypothesis.
- Morphological complexity
The response of subjects may vary significantly based on the morphological complexity of the data set; therefore, the data set should be chosen very carefully to reflect the purpose of the study. Further, if more than one datasets are being used contrastively, then their morphological complexity should be identical.
- Number of sounds/words in a word/sentence
This may seem a relatively unimportant factor. However, the studies have shown that these affect the performance of subjects during the experiment. So this variable should also be kept in mind while designing an experiment.
- Effect on results
There is no concrete basis to determine the effect of these problems on the findings from an experiment; however one thing is sure that involvement of one or more problems may increase the possibility of deviation of findings a lot from the accurate findings, which in turn may result into failure of a hypothesis or formation of a hypothesis on wrong premises. Further, the size of the sample and the number of subjects involved also affect the probability of deviation from accurate finding.
Successful execution and getting uninfluenced results from an experiment are highly dependent on the design of the experiment. Therefore, a well-worked experiment design can be very helpful in getting the desired results and saving lots of rework. Artificial Intelligence clubbed with big data is solving various herculean problems effectively in the healthcare sector these days. Therefore, development of such tools which may predict the vulnerability factor of a test based on various constraints could be a significant milestone in this direction
- Alfonso Caramazza, Michael McCloskey. (1980) Theory And Problems in Psycholinguistics. Annual Review of Applied Linguistics 1, 71.
- Daniel C. O’Connell, Sabine Kowal. (2011) Sources of History for “A Psychology of Verbal Communication”. Journal of Psycholinguistic Research 40:1, 29-47.
- James W. Goodwin, Uwe Hein. (1982) Artificial intelligence and the study of language. Journal of Pragmatics 6:3-4, 241-280.
- Joel Walters, Yuval Wolf. (1986) LANGUAGE PROFICIENCY, TEXT CONTENT AND ORDER EFFECTS IN NARRATIVE RECALL. Language Learning 36:1, 47-64.
- Joseph F. Kess. (1991) On the developing history of psycholinguistics. Language Sciences 13:1, 1-20.
- Rainer Dietrich, Carl F. Graumann. 1989. Language Processing in Social Context. An Interdisciplinary Account. Language Processing in Social Context, 1-15.
A sample self-paced reading experiment design for Hindi Relative clause
- 1. To find if there is any difference in the neuro-cognitive resources used in cognition of Relative clause and a simple structure of a semantically equivalent utterance.
- To find out the time required for comprehension in both cases.
(a) Data: Vo baccā jo vahāṃ khaɽā hai, mere sātha jānā cāhatā hai.
(b) Contrastive data: Vahāṃ khaɽā huā baccā mere sātha jānā cāhatā hai.
(Dwivedi, Satyam. “Psycholinguistic Experiments: Design, Potential problems and Addressing them.” Current Journal, V, no. 17, January 2018, pp. 127–130.)
PC: Unknown illustrator