http://clive-shepherd.blogspot.com/2007/04/e-learning-and-science-of-instruction.html

Use of media

  • Use words and graphics rather than words alone (89% gain in learning).
  • Keep graphics and text that relate to each other near each other (68% gain).
  • Where possible, describe graphics using audio narration rather than text (80% gain). An exception here would be text (unfamiliar terms, instructions, etc.) which require time to process.
  • Avoid presenting words as both narration and text (79% gain).
  • Cut out extraneous/non-essential text, audio and graphics (82% gain).
  • Use a conversational style, using the first and second person, for both text and audio (67% gain).

Practice questions and worked examples

  • Better learning results when practice questions are distributed throughout the learning, rather than all at the end.
  • Questions that ask the learner to merely recognise or recall information previously provided in the training will not promote learning that trasfers to the job.
  • Transfer is maximised when the practice questions mirror real-work situations.
  • For critical tasks, such as those with safety consequences, more practice is required.
  • The more practice the better the learning.
  • Instructions for practice questions and feedback should be presented as text rather than audio and placed alongside the question.
  • Where audio or video is needed for practice, include a replay option.
  • Worked examples/demos are popular with learners and can replace some practice questions.
  • For procedural tasks, a single worked example is likely to be adequate.
  • For problem-solving tasks, a wide range of worked examples might be needed.

Collaboration

  • Collaborative tasks should be designed in such a way that they require learners to interact and contribute to a group outcome, i.e. they cannot be achieved by single participants working alone.
  • Collaborative tasks work best with learners working in pairs or groups of no more than six.
  • Heterogeneous groups get better learning outcomes than homogeneous groups.

Control

  • Learners like learner rather than program control.
  • Student preferences and judgements often may not be good indicators of the way they learn best.
  • Use learner control for learners with high prior knowledge or metacognitive skills and/or for courses that are advanced rather than introductory.
  • When using learner control, design the default navigation options to lead to the most important instructional elements.
  • Make sparing use of links that take the learner away from the current screen or which provide the primary means of access to important elements of the course (because most learners will regard these as peripheral).
  • Provide advice to help learners make decisions about what to do next.
  • Use program control when most of the audience is likely to be novice and/or high levels of skill attainment are critical.
  • Use course maps to provide an overview and orient learners.
  • Provide basic navigation options (back, forwards, menu, exit) from every display.

Clive on Learning: e-Learning and the Science of Instruction