Web-Based Language Testing

Web-Based Language Testing: Opportunities and Challenges

Computer-based testing

computer-based tests have been used in SL testing since the early 80s, as a way to deliver traditional tests via the computer or even adaptive tests based on item-response theory
compared to paper-and-pencil tests, CBTs offer logistical advantages, a unified platform for media, diverse item types, tracing of test-taker actions, and immediate feedback
the high frequency of test administrations leads to significant problems with item security
writing CBTs requires high levels of programming expertise ð they tend to be very expensive to produce and maintain

Web-based testing

a web-based test (WBT) is a computer-based test delivered via the internet and written in the "language" of the internet, HTML and possibly enhanced by scripts
the test is located as a website on the tester's server where it can be accessed by the test-taker's computer, the client. The client's browser software (Netscape Navigator, MS Internet Explorer) displays the test, the test-taker completes it and—if so desired—sends his/her answers back to the server, from which the tester can download and score them
if the test consists of dichotomous items (T/F or MC) it can be made self-scoring by using scripts.

Advantages of WBTs:

Principle of "anyplace, anytime": anybody with WWW access and a browser can complete the test at any time from anywhere in the world. This is not the case for traditional CBTs which need special delivery platforms.
"Testing goes grassroots": WBTs are very easy to write, anybody with basic knowledge of HTML can write a simple but professional looking WBT, even using images, sound and video. Anybody with knowledge of scripts can write a sophisticated WBT, which can be self-scoring or even adaptive. Neither is the case for traditional CBTs which require programming expertise.
"Testing goes affordable": WBTs are very cheap to produce and maintain. Production requires a computer, a web editor (comes as a part of a free Netscape Communicator or MS Internet Explorer package) and an HTML reference book. To put the test on the web, your institutions will probably give you free web space or you can put the test on a free site on the net (for example, www.geocities.com or www.tripod.com)
"Testing goes immediate and scorer-free": for any dichotomous test (one that has only right/wrong answers), the scoring can be handled by a script, relieving testers of the tedious task of scoring and providing test takers with immediate feedback on their performance even on an item-by-item basis if so desired. Scripts are not as easy and intuitive as HTML but learnable with a good reference book and some practice.

Problems of WBTs:

Identification: Are test-takers really who they say they are? How do you make sure that your test takers did not have outside help, used a dictionary, asked a friend?
You can't. No technology can assure standardization of testing conditions if people are taking the test in the privacy of their own home. So WBTs should be restricted to low-stakes tests or tests where it's in the test-takers best interest not to cheat, e.g., self-assessment instruments. Of course, if testing is conducted at testing centers under supervision, security problems can be minimized but that would defeat the "anytime, anyplace" principle of WBT.
"A test's web appropriateness is inversely proportionate to its stakes."
Item Confidentiality: There is no way to prevent item downloading and subsequent exposure without compromising the "anyplace, anytime" principle by supervising test administrations. Only use disclosed, non-confidential items.
Secure Data Storage: test taker responses must be automatically stored on the server or a disk in case of system failure. Disk storage is difficult to realize technologically.
Secure Data Transfer: for high stakes tests data must be protected against interception and should be encrypted.

Types and characteristics of WBTs by stakes involved:

Stakes	Security Need	Web Advantages	Cost	Example
low	little	anytime, anyplace	low	feedback, test-prep, independent tasks, "testing in the service of learning"

medium	ID, general supervision, data backup	almost anyplace during opening hours	some personnel overhead	placement testing, course credit

high	tight supervision, controlled environment, multiple backups	only trusted & certified testing centers	probably high fees, costs for backup	TOEFL-type tests, admission tests

Some usage contexts for WBTs in order of web appropriateness:

self-assessment / assessment in the service of learning: test takers receive feedback on their performance, areas that need more practice can be identified, testers do not have to be involved in scoring. Low stakes, test takers have no vested interest in distorting their true scores.
research: participants can take test at their convenience, have no incentive to cheat.
achievement testing: for example to earn credit for a distance education course. Requires supervision but need for item confidentiality is low for items based on real-world tasks.
placement (post-admission): no more centralized placement tests necessary, institutions can plan course capacities early. Requires supervision and item confidentiality.
large-scale, high stakes, pre-admission proficiency testing: almost no advantages of WBT over traditional CBT except easier production but strict supervision required, strict item confidentiality, high level of data security. Not recommended for WBT at this time.

WBTs and self-guided learning

WBTs are ideal for web-based supplementary, independent after-class activities because
different students can be assigned different activities, depending on need,
students can work at their own pace,
scripts can provide immediate feedback and crude adaptivity (by means of pre-set break-off criteria),
teachers can choose their degree of involvement, from none to minimal (score and time reporting) to maximal (detailed reports of responses)

Item types for WBTs and their web realizations:

Item	Realization and Scoring	Example
essay	textarea, requires human scorer	extresp.htm
multiple choice	radio buttons, easy to write in HTML; view the scoring script for this test (external .js file)	if.htm
short answer	input fields; automatic load of the next items into top frame; to view the scoring script, pull up bottom frame border, click the right mouse button and view "Page Source".	rela2.htm

gaps (grammar, Cloze, C-Test)	input fields; to view the scoring script for this test, view "Page Source"	cucar.htm
oral production	test takers record .wav file with sound recorder, requires human scorer; technical problems likely	---
dictation	sound file plus textarea, self-scoring is possible but very difficult	dictfr.htm
listening comprehension	sound file(s) plus radio buttons, input fields or textareas, any response format possible	---
discourse completion	situation described in writing, in a sound file or shown in a video clip, any response format possible	salad_bowl.htm

Resources:

Röver, C. (2000). Web-based language testing: A manual. Manuscript in preparation.
Vasu Renaganathan's WebAuthor: http://ccat.sas.upenn.edu/plc/larrc (some bugs)
Hot Potatoes: http://web.uvic.ca/hrd/halfbaked/

Pfaffenberger, B. & Gutzman, A.D. (1998). HTML 4 Bible. Foster City, CA: IDG Books.
Goodman, D. (1998). JavaScript Bible. Foster City, CA: IDG Books.

Chalhoub-Deville, M. & Deville, C. (1999). Computer adaptive testing in second language contexts. Annual Review of Applied Linguistics 19, 273-299.
Brown, J.D. (1997). Computers in language testing: Present research and some future directions. Language Learning & Technology 1 (1), 44-59. Available: http://polyglot.cal.msu.edu/llt/vol1num1/brown/default.html
Dempster, F.N. (1997). Using tests to promote classroom learning. In Ronna F. Dillon (Ed.), Handbook on testing (pp. 332-346). Westport, CT: Greenwood Press.