As many of you probably know, almost all advice regarding doing better academic talks includes universal advice to not put too many words in any one slide, and to not read straight from slides (this is always thought to be a sign of a bad talk).
I used to agree with all of this because it sounds reasonable, and anyway that's how I've written my own talks so far. But I went to a conference late last year where a speaker did exactly the opposite of this advice (my field is mathematics). He was literally reading his slides verbatim, and each of his slides was packed full with long sentences. And yet, it was one of my favorite talks of the conference, and for me (a Ph.D student) it was clean, clear, and easy to follow despite the fact that the material was completely new to me. I do think a large part of the reason I found it so clean and clear was precisely because of the talk's structure, and not despite of it.
That experience made me rethink my prejudice against speakers who read out of the slides. For some people, it may be a superior talk strategy to alternatives, especially in academic fields where the details really matter.
So, I am not convinced anymore that it is general (or even, usual) good advice to not read from the slides. Why do people think it's such a bad idea? And please don't say "it's lazy" or "could just read the slides instead of listening to the talk" because in practice, neither of these perspectives demonstrate why an alternative is better at communicating the information, which is what really matters.
Research in the field of multimedia learning suggests that in most cases, "people learn more deeply from graphics and narration than from graphics, narration, and onscreen text". This is known as the redundancy principle.
Previous research has shown students learn better from multimedia lessons containing graphics and narration than from graphics, narration, and redundant on-screen text (Kalyuga, Chandler, & Sweller, 1999, 2000, 2004; Leahy, Chandler, & Sweller, 2003; Mayer, Heiser, & Lonn, 2001; Moreno & Mayer, 2002a, 2002b; Mousavi, Low, & Sweller, 1995). This finding is known as the redundancy effect (Mayer, 2001, 2005c). For example, in a study by Moreno and Mayer (2002b), participants viewed an animation about lightning formation. The first condition had narration accompany the animation, whereas a second condition received redundant on-screen text in addition to the animation and narration. The group that received the redundant on-screen text performed worse on subsequent retention and transfer questions than did the group that received animation and narration; thus, a redundancy effect was found.
Adding redundant on-screen text detracts from the learning processes highlighted in Figure 1 because it creates extraneous processing—such as inducing the learner to visually scan between the caption at the bottom of the screen and the graphic and to try to mentally reconcile the incoming spoken and verbal stream. If the learner has to waste limited cognitive capacity on extraneous processing, the learner will be less able to engage in the cognitive processing needed for learning—essential and generative processing.
(also from )
Are there times when it is good to include some of the narration in a slide? Yes. For example:
The most straightforward contribution of this work is to add an important caveat to the redundancy principle “People learn more deeply from graphics and narration than from graphics, narration, and onscreen text” (Mayer, 2005c, p. 193). In revising this statement of the redundancy principle, we can add the following limitation: “except when the on-screen text is short, highlights the key action described in the narration, and is placed next to the portion of the graphic that it describes.”
In a meta-analysis of the literature on the redundancy principle:
[A]dvantages of spoken–written presentations over spoken-only presentations were found for low prior knowledge learners, system-paced learning materials, and picture-free materials. In comparison with verbatim, spoken–written presentations, presentations displaying key terms extracted from spoken narrations were associated with better learning outcomes and accounted for much of the advantage of spoken–written over spoken-only presentations.
You perceived the talk with redundant written and spoken text to be beneficial to you; this is consistent with the research as well:
Research on metacognition has consistently demonstrated that learners fail to endorse instructional designs that produce benefits to memory, and often prefer designs that actually impair comprehension. Unlike previous studies in which learners were only exposed to a single multimedia design, the current study used a within–subjects approach to examine whether exposure to both redundant text and non-redundant text multimedia presentations improved learners' metacognitive judgments about presentation styles that promote better understanding. A redundant text multimedia presentation containing narration paired with verbatim on–screen text (Redundant) was contrasted with two non-redundant text multimedia presentations: (1) narration paired with images and minimal text (Complementary) or (2) narration paired with minimal text (Sparse). Learners watched presentation pairs of either Redundant + Complementary, or Redundant + Sparse. Results demonstrate that Complementary and Sparse presentations produced highest overall performance on the final comprehension assessment, but the Redundant presentation produced highest perceived understanding and engagement ratings. These findings suggest that learners misperceive the benefits of redundant text, even after direct exposure to a non-redundant, effective presentation.
Note that the results above refer to multimedia presentations, i.e. those that include graphics. Presentations that do not include graphics may benefit from redundant spoken-written text (see ). But presentations that do not use graphics violate the multimedia principle, which states that in general, "people learn better from words and pictures than from words alone".
For more information on principles of multimedia learning, see The Cambridge Handbook of Multimedia Learning. If you're looking for something shorter, here is a brief summary of the major principles of multimedia learning, and here is a presentation by Richard Mayer.
 Revising the redundancy principle in multimedia learning. Mayer, Richard E.; Johnson, Cheryl I. Journal of Educational Psychology, Vol 100(2), May 2008, 380-386. http://dx.doi.org/10.1037/0022-0622.214.171.1240
 Verbal redundancy in multimedia learning environments: A meta-analysis. Adesope, Olusola O.; Nesbit, John C. Journal of Educational Psychology, Vol 104(1), Feb 2012, 250-263. http://dx.doi.org/10.1037/a0026147
 Learners misperceive the benefits of redundant text in multimedia learning . Fenesi, Barbara; Kim, Joseph A. Frontiers in psychology 5, 2014. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4088922/
 Verbal redundancy in multimedia learning: When reading helps listening. Moreno, Roxana; Mayer, Richard E. Journal of Educational Psychology, Vol 94(1), Mar 2002, 156-163. http://dx.doi.org/10.1037/0022-06126.96.36.199