Devenagari Script and Reading.

‘Segmental Discrepancy Between Script and Reading’
Experimental Findings on Devenagari


Introduction:
There might exist difference(s) between the way a language is written and the way it is pronounced. The variation between two realizations, at orthographic and articulatory levels, has often created the most intriguing debates. This discrepancy between script and sound of a language is observed the best in the events like reading of texts. It is believed that some characters and their resultant word forms will be found more difficult in processing than others. This might, also, vary from language to language, as some languages are alphabetic, some syllabic and some are pictorial.
The Indian languages have also been found to contain such discrepancies between word forms and sound forms. Although the characters in Indian scripts are moreover phonemic in nature, scholars like Sproat, 2006 argue that the Brahmi derived Indian scripts creates a sort of inconsistency between the written and articulated of a certain syllables types. Thus, causing a delay of inaccuracy in the processing of those syllables. They include, for example, in Devanagari script (as used in Hindi) the high front vowels short [i] and long [i:], in words like ‘kitaab’ meaning book and ‘jiivan’ meaning life respectively. According to Sproat, 2006 there is scriptal influence on the phonemic awareness in readers of Indic scripts. Whereas, such influence can be minimized if the script supports phonemic operation in certain environment.
The aim of this paper is to present some experimental findings with respect to segmental processing of ‘Devenagari’ script and present a case against Sproat’s Edit Distance Hypothesis which says the transformation form graphic form of the stimulus to the graphic form of the response. That is to say the discrepancy between orthography and pronunciation of the segments does not result in inadequacy or delay in lexical processing. It’s just convention.

Method
Materials:
The items list included of twenty pseudo words of two syllables each. There were arranged in such a way that each of the four conditions under which the two foresaid vowels may appear was displayed for five times during the whole test. The conditions were:
Condition One: VCVC structure
Condition Two: CVCV structure
Condition Three: VCCV structure
Condition Four: CVVC structure
Both vowels preceding: As in a non-words like ‘pimi’,’kigi’etc.
Both vowels following: As in non-words like ‘piimii’, ‘giikii’ etc.
First preceding and second following: Like, ‘rijii’, ‘pikii’ etc and
First following and second preceding: Like, ‘giipi’, ‘liini’ etc.
Formally summarized as VCVC, CVCV, VCCV and CVVC. Also, there were thirty more pseudo words that looked exactly like the items but were indeed fillers to distract the subject and prevent them from guessing. The words would appear one by one for a set time duration and disappear thereafter. Next there would appear four options out of which only one was true. Once the choice is made the next word would appear on the screen and so on.


Participants:
The participants included 14 native speakers of Hindi language. They aged between 18 and 40 and 12 of them were right hand oriented while the remaining 2 were left hand oriented. The subjects comprising of males and females both, were given clear instructions on what they were supposed to do at the start of the experiment as well as they were given a test run so as to get familiarize with the experiment. It was a voluntary participation from the subjects and no money was involved. They were chosen on uniform criteria and were experimented on the same computer system under the same environment.

Procedures:
The test comprised of 50 pseudo - words display against which centred auto paced reading task was to be performed. Of the 50 non-words twenty were intended items that represented the four conditions viz VCVC, CVCV, VCCV and CVVC. Also, there were thirty more non words that resembled the items but were fillers. The fillers were used just to distract the subject from predicting the items and to maintain the reliability of the test.
1.At first a word would appear for 80-100 Msec in the fixated centre of the screen.
2.And it would disappear after that, leaving the question what did you see?
3.Then there would appear four options out of which one was the displayed word.
4.The subject would press button to decide the word s/he has seen.
5.The display was centred and auto- paced.
6.After a word was chosen from the list another word would appear on the centre of the screen.
7.The time taken in decision making and the accuracy were recorded for computation.
8.During computation incorrect response and the responses to the fillers were not considered.
For the experiment, which involved the use of computer, we used a programme called Linger. A brief introduction to the programme used in the experiment is as under:
Linger is a Tcl/TK application that can be used to perform a variety of language experiments. It can be modified easily and used on all major operating systems like Windows, Linux, Unix and Mac. It can be used with just about any language having left to right horizontal script for which a font is available. Linger can handle a variety of experimental techniques like true/false, multiple choice questions, fill in the blanks, and difficulty ratings, as well as audio playing and recording.

The experiment that can be performed on Linger include:
Masked Self –Paced Reading
Centred Self Paced reading/ Listening
Masked Auto –Paced (fixed rate) Reading/ Listening
Centred Auto – Paced Reading/ Listening
Block Reading
Speak and Listen
Listen and Answer
Auditory Prime Lexical Decision
The choice of the programme Linger was motivated by the following factors:
Specificity: It is specific to computing language experiments.
Sensitivity: It shows up reliable differences for intended results.
Controllability: It matches materials adequately.
Availability: It’s available freely.
The experiment is basically a centred auto paced reading task. It is called centred because, the target stimulus non-word would appear in the centre of the computer screen. It’s called auto paced for the reason that the duration of display would be fixed in advance, meaning that the word won’t stay on screen forever. Here, the display was in the range of 80 -100 msec. The display was followed by four similar choices out of which only one would be right. The subjects were expected to recall that word they had seen and chose the same from the choices. Once done they would press the corresponding button. This would also lead them to the next word. The display of the target stimuli and the fillers was randomized through the programme, so that there is no scope for prediction and the bias on the part of the subjects could be minimized.

Results and Discussion
The time set for display 80 Msec was adequate for the subjects to pick the word [non-word]. So, keeping it in the range of 80-100 msec ensured 85% accuracy with similar time taken for decision. However, the gender, handedness and age did not seem to play a role in this task. We chose non words as target stimuli because the processing of meaningful words can be motivated by several other factors like template matching etc. The response analysis shows that the difference between syllabic form in script and reading does affect the processing in terms of response time or accuracy. So, the transformational delay [Edit Distance Hypothesis] as proposed by Richard Sproat is contradicted. We may conclude that the native or nativized speakers or readers tend to adopt the discrepancy as convention and are quite apt in their own script. There does not seem to be any difference in the difficulty or ease if the script is changed.
Reference
Sproat R et al. 2006, Scriptal Influence On Phonemic Awareness In Readers Of Indic Script.