MARIA KOUTSOUBOU
3. THE DESIGN OF THE STUDY 1 The Tasks
Twenty Greek deaf students participated in the study. The two variables considered were the stimulus material and the bilingual language abilities. Two sets of stimulus materials were designed. The first was a story presented on video in Greek Sign Language. The second was a picture storybook without printed text. In both tasks the requirement was to write the story down. The aim was to compare the stories elicited by the different material and to decide which was more elaborated in infor- mation, organization and language use.
These specific tasks were chosen because they replicate features of either a bi- lingual approach in the classroom (video) or a traditional approach to teaching deaf children (picture book). In the video task, sign language is explicitly involved in the writing process, as it is a translation task, In the picture book task, there is no ex- plicit source language involved in writing. The video task may therefore be expected to show more interference from sign language. If similar errors are found in the pic- ture task, it may indicate that in both situations sign language is used to create mean- ing and to facilitate the process of writing in Greek.
3.2 Participants
The second variable – bilingual language abilities – was determined by assessing the two languages involved in the writing process i.e., Greek Sign Language (GSL) and written Greek. Although several sign language assessment tests have been devel- oped for British Sign Language (Herman, Holmes, & Woll, 1999) and American Sign Language (ASL) (Mounty, 1993) there are no standardized assessments for GSL. The researcher therefore constructed a scale with four levels, each defined in terms of general linguistic and pragmatic characteristics. Teachers were asked to rate each student’s performance in GSL and written Greek. For a full description of the assessment see Koutsoubou, in preparation. Deaf teachers assessed the GSL and deaf and hearing teachers assessed written Greek. The assessment procedure re- sulted in the division of the sample into three groups:
1) 2) 3)
Sign Language Dominant group (GSL high, written Greek low) 6 subjects.
Weak balanced bilingual (GSL low, written Greek low) 6 subjects.
Strong balanced bilingual group (GSL high, written Greek high) 8 subjects.
3.3 Materials
The materials used, were two picture stories without words of about the same length.
The “Frog, Where are you?” (Mercer, 1969) and the “The Grey Lady and the Straw- berry Snatcher” (Bang, 1986). Both were signed by a deaf native signer of GSL. The stories were piloted with a bilingual GSL and Greek hearing subject and both stories elicited a similar length and a similar degree of grammatical complexity. Half of the participants received the Frog Story in GSL and the Strawberry Lady in pictures and the other half received them the other way round, in order to control for story ef- fects.
3.4 The Analysis of Texts
The texts have been analyzed in four levels:
Information (amount and type of information revealed) (see appendix A). This is measured in two ways: (a) the basic structure of the story. This consists of the setting, the reason, the action, and the closure. This is applied to both stories.
The terminology used is the researcher’s but the approach is based on previous research (Bamberg, 1997; Labov, 1997; Yoshinaga-Itano & Downey, 1998); (b) the basic story lines, which are specific to each story.
The type of information is also measured in two ways: (a) descriptive informa- tion – any state or action in the story; (b) affective information, any information about the thoughts of the characters, dialogues, monologues, comments, attrib- utes, or evaluations by the narrative writer.
Organisation (how the information is structured) (see appendix B). This is measured through the use of tree diagrams (Langer, 1986). The deeper the tree grows, the tighter the structure of the story. The wider the tree is, the more in- formation is present. Finally the greater the variety of relationships found in the tree, the better elaborated the story is. Langer’s original model has been modi- fied to some extent, since the language produced here is quite different – due to its bilingual nature – from that for which the tree diagrams were designed.
Text characteristics. These are standard measurements of the complexity and well-formedness of written language. The measures used here are: number of words, number of T-Units, number of clauses, clauses per T-Units, T-Unit length, Subordinate clauses, Co-ordinate clauses and T-Unit complexity (Fraser, 2001; Silliman et al., 2000).
Error analysis (see appendix C) (James, 1998). This is focused on language form, with a particular focus on the weaker parts of the texts
1)
2)
3)
4)
3.5 The Hypotheses Tested
Comparisons between the three groups and within each group have been made to see whether there are differences in the information level, organization level and gram- matical level between the tasks and among the groups. There are two hypotheses:
1) The different bilingual groups will produce different texts in quality and quan- tity with different characteristics in organisation, grammar and information.
The picture and video material will produce different texts in quality and quan- tity of organisation, grammar and information revealed.
2)
4. RESULTS
Data were analyzed using general linear model-repeated measures with SPSS. The model used was repeated measures. The within-variables were the video and picture material (two levels) and the between-variables were the three groups (three levels).
The graphs are boxplots, which show the median (the thick black line), the &
the percentiles of the values and the largest or smallest values (indicated by the whiskers). Small circles also indicate outlying and atypical values. The level of sig- nificance was set on 0.05 for all the analysis in this study.
4.1 Main Effect of Groups
The strong balanced bilingual group (the group defined as having good written Greek skills) performed consistently and significantly better than the other two groups in most of the measures. Of interest, however, are the differences between the SL dominant and the weak bilingual group as well as their relation to the strong bilingual group.
4.2 Results: Information Level
On the story grammar information the only significant difference between the groups was between the strong-balanced and the weak balanced group (p = .045) with a main effect of F (2, 14)= 4.784, p = .026 (see Table 1).
The SL dominant group does not differ significantly either from the bilingual group or the weak balanced group. This makes the SL dominant group a middle group, representing bridge between the high achieving strong bilingual and the low achiev- ing weak bilingual group. Figure 1 presents data on the groups in greater detail. The strong balanced group performs better in both conditions than either of the other two groups. The weak balanced group produces the lowest scores in both tasks and the SL dominant group is located in the middle – although it should be noted that this
group exhibits more variability in scores. This distribution is also found in other analyses.
Figure 1. Performance of the three groups on story grammar production in the video and picture task.
On the basic story line information the main effect of group is F (2, 14)= 7.570, p = .006 and the pairwise comparison between the strong-balanced group and weak- balanced group was significant (p = .008). Again the SL group’s scores are between these two groups with no significant difference from either (Table 2). Figure 2 shows the wide distribution of scores within the SL dominant group.
Figure 2. Performance of the three groups on basic story line production in the video and picture task.
4.3 Results for Analyses of Organization
On the variety of relations produced in narratives the main effect of groups is F (2, 14) = 6.646, p = .009 and the pairwise comparison between the strong-balanced and weak balanced groups was significant (p = .014). In this analysis, the SL dominant group more closely resembled the weak balanced group than the strong balanced group, although results were not significant (Table 3). This can be seen more clearly in Figure 3 where the medians of the two groups are almost on the same level.
Figure 3 shows that the weak balanced and the SL dominant groups perform at a similar level, although they react differently to the material: writing from video is better for the SL dominant group and writing from pictures is better for the weak bilingual group.
Figure 3. Performance of the three groups on the variety of relationships in the video and picture task.
4.4 Results for Analyses of Text Characteristics
On this measure, the strong bilingual group performed significantly better than ei- ther of the other groups. The text characteristics were similar for the weak balanced and the SL dominant groups.
4.5 Results for Error Analysis
There were no significance differences on error analysis between any of the groups, with all groups having approximately the same amount and type of errors.
4.6 Main Effect of Material
There was a main effect of the material on the organization level and the error analy- sis. The relevant results for organization of the stories relate to the and level of the tree diagrams (see appendix B). The video produced significantly better re- sults than the picture book on the level [F (2, 14) = 7.363, p = 0.017] (see figure 4) and the level of the story diagrams [F (1, 14) = 5.924, p = 0.029] (see Figure 5).
Figure 4. Performance of the three groups on the level of story organisation within the video and picture task.
Figure 5. Performance of the three groups on the level of story organization within the video and picture task.
As for error analysis (see Appendix C) the type of stimulus material produced only one effect that was close to significance: the “grammar - omission” category [F (1, 14) = 4.348, p = .056]. “Grammar-omission” means that the writers omitted various grammatical words from the text. The video produced more omission errors in the
strong balanced and the SL dominant but there was no effect on the weak balanced group (see Figure 6).
Figure 6. Performance of the three groups on errors of omission within the video and picture task.
Some of the omission subcategories yielded results approaching significance:
“omission of preposition” [F (1, 14) = 4.178, p = .060] (see Figure 7), and “omission of verb” [F(1, 14) = 5.149, p = .040] (see Figure 8). More omissions were found in the video condition than in the picture condition. These finding are in accordance with previous studies of the writing of deaf students. Deaf students frequently omit- ted many function words and verbs. Most often state and existence verbs are omitted (e.g., “to be”, “to have”, “to appear”). This is probably due to sign language interfer- ence, as these omissions are noted only in the two groups with good sign language skills: strong balanced bilingual and SL dominant. The scores for the weak balanced group, although showing a tendency in the same direction (i.e., producing more omissions in the video task than the picture) were not significantly different from the picture task (see Figure 6 and 7). Verbs were significantly omitted from all the groups (see Figure 8).
Figure 7. Performance of the three groups on errors of omitting prepositions within the video and picture tasks.
Figure 8. Performance of the three groups on errors of omitting verbs within the video and picture tasks.
4.7 Interaction Effect between Group and Material
The only significant difference of interaction between groups and material occurred at the information level – affective type [F (2, 14) = 4.124, p = 0.039]. The video significantly improved the strong balanced bilinguals’ performance, significantly
impaired the weak balanced bilinguals’ performance and had no significant effect on SL-dominant performance, as seen in Figure 9.
Figure 9. Performance of the three groups on the production of affective information within the video and picture task.
5. DISCUSSION
Writing is a complex activity, which starts with decisions about how much informa- tion and what type of information a text should have, progresses to organizing the information and finally to decisions about how everything will be linguistically ex- pressed. The type of input – in the present case, sign language input – facilitated different aspects of writing and interacted in different ways with various levels of bilingualism.
At the higher levels of writing (i.e., decisions on information and organization of the story) the availability of sign language and language fluency have a significant effect on deaf students’ writing. The strong balanced bilingual group improved its performance in the sign language condition, whereas the weak balanced bilingual group and the SL Dominant group showed more limited effects. In general, sign language source material improved the structure of the texts in terms of organization (Figures 3, 4 and 5). In relation to text characteristics, the source material caused no significant effect. The only negative effect of sign language was found in the error analysis and that occurred only in omission of function words such as prepositions, and verbs of state/being. This can be interpreted as an effect of transfer from one linguistic system to the other. It has also been described in the literature as one of the characteristics of deaf writing (Paul, 2001). Apart from omissions, the two types of materials produced text of more or less the same quality. This is evidence that the language used in thinking for writing in the context of the pictures was similar to that used in the translation (video task): Greek Sign Language.
The effect of sign language on writing demands attention to the issue of how to use it most effectively in schools. Even if sign language has been accepted as a language for use in deaf education, deaf students’ sign language skills are not formally as- sessed or even scrutinized. Deaf students should be treated as bilinguals with vary- ing skills in sign language. Such an approach may lead to consideration of grouping deaf children in classes according to their language skills and not according to their age. This of course will only be fully feasible when assessment tests for sign lan- guage are developed and standardized.
A second conclusion from the results is that teachers need to consider what types of source materials can be used in order to improve specific aspects of deaf students’
writing. A third conclusion is that error analysis provides the only evidence of sign language material having a negative effect on all groups. In all the other levels of analysis, the strong balanced bilingual group significantly outperformed the other two groups. Even though errors are present, the texts with the greatest number of errors are not necessarily the worst texts. Error counting is a fairly low level of analysis, and meaning can be passed on even in the absence of correct grammatical form. What makes for a good text is the provision of all necessary information, good organization and good discourse manipulation. These were best with the sign lan- guage source material.
In conclusion, this study has provided us with answers to the questions posed at the beginning:
Do deaf writers’ of different bilingual experiences make different use of the linguistic input? With the exception of one result (differences on affective in- formation, see Figure 1), the groups do not differ very much in terms of use of linguistic input.
Can we influence the process of writing by using different material? We can influence some aspects of writing but not all. For example by using sign lan- guage, we can influence information and organisation of written stories but not necessarily the text characteristics.
Do the patterns of errors change when we change material or do deaf students always go via the same route? Since – as indicated by the answer to Question 1 – the groups do not react differently to the materials, they are likely to be using them in more or less the same way. With the exception of omission of words, both materials produce the same kinds of errors. It is obvious that the video task – as a translation task- was more vulnerable to sign language interference. But it may be concluded that the picture task – which was free from language input – produced similar types of errors. This may indicate that deaf students, regard- less of their sign language skills, resort to sign language to form their texts.
1)
2)
3)
One final comment may be made on the allocation of groups. The fact that the SL dominant group is always placed between the best performers and the weakest per- formers indicates that the development of a high degree of fluency in a sign lan- guage can facilitate literacy in deaf students.
APPENDIX A
1The table for “The Strawberry Lady ” is given as an example in the Appendix. The same structure was applied for “The frog Story”.
APPENDIX B
FRAGMENT OF A TREE DIAGRAM FOR STORY ORGANISATION
This is a fragment of a tree diagram, indicating the structural relationships in a story.
Four levels are represented, the deepest has three clauses (clause number 19, 22, 24) and there are four different types of relationships presented (Event/Description/ Ad- versative/Explanation). The numbers under the relationships refer to clauses in the narrative. The written text it refers to and the clause segmentation are following:
[...] Some other time again he saw a lady who has the strawberries was running and followed but lady disappeared in the wood But is boy disap- pointing because not is-found the strawberries […]
Clause segmentation:
18) 19) 20) 21) 22) 23) 24)
= Some other time again he saw a lady
= who has the strawberries
= was running
= and followed
= but lady disappeared in the wood
= But is boy disappointing
= because not is-found the strawberries
APPENDIX C
This page intentionally blank
Eds.). Studies in writing. Vol. 14, Effective learning and teaching of writing, Edition, Part 1, Studies in learning to write, 169 - 179.