Power Point based on the article "Testing for language teachers" (Arthur Hughes), pages 83 to 112 (Chapter 9: Testing writing). This work is done by Idoia Argudo and Marta Ribas, in a subject from Universidad de Cantabria.
2. How to test people’s writing?
Get them to write
Test the ability directly
• Tasks properly representative
• Tasks that represent students’ ability
• Scored validly and reliably
3. Representative task:
specify all posible content
Be clear
• Operation: describe, explain and compare.
• Types of text: examination answers up to two paragraphs in length.
• Addresses of text: native speaker and non-native speakers university
lecturers.
• Topics: academic treatment, not specialist.
• Dialect and style: any standard variety of English. Formal style.
• Length of the text: about 1 page.
Example:
4. Representative task:
include a representative sample
Set a task where the students can prove their truly
ability.
The more tasks, the more representative.
Wide range of representative specifications →
backwash effect.
8. The more tasks we set the more
representative/valid of a
candidate’s ability we will obtain
Spend too much time
9. Representative task:
include a representative sample
If the test is used
to
Place students in
classrooms
Not so important
accuracy
Determine if they
are allowed to
study overseas
More than one
sample is
needed
Topics should be familiar to the students.
10. Elicit a valid sample:
set as many separate tasks as possible
Desirable
Practical
Offer as many
fresh starts as
possible
Achieve greater
reliability and
validity
11. Elicit a valid sample:
test only writing ability
Creative, imaginative, intelligent, wide general knowledge, good reasons for
the opinions they hold…
Test ONLY writing ability
For the sake of validity:
12. Elicit a valid sample:
test only writing ability
To not infere with Reading ability, we
should use simple instructions or
illustrations.
Candidates should know what is
required from them.
Tasks as authentic as possible.
14. Ensure valid and reliable scoring
Set tasks which
can be reliably
scored
Set as many
tasks as
possible
Restrict
candidates
Give no choice
tasks
Ensure long
enough samples
Create
appropiate
scales for
scoring
15. Ensure valid and reliable scoring
Holistic scoring Analytic scoring
Single score of a writing (overall
impression of it)
Separate score for each aspect
Assess student work as a whole Assess the components of a finish
product
Fast to correct Slow to correct
General test Diagnostic information
Small, well-knit group Heterogeneous, less well trained
group
Example: TOEFL Used generally in schools
17. Conclusion
• If the test includes a wide ranging and representative sample of
specifications, the test is more likely to have a beneficial washback
effect.
• A test that required the student to write 4 answers could cover the
whole range of tasks.
• Chose the topics that students are familiar with.
• Provide the precise information and arguments.
• Adapt the writing to the students level.