He cites all well-known and not at all surprising problems with peer review, which means that a sizeable proportion of published science is wrong -- wrong as in somewhat innaccurate and likely perfectible, and in now way as wrong as only religious ideas can be, i.e. completely false and unfounded. Asimov wrote a great essay about this Relativity of Wrong [1].
Indeed, that fact that science isn't regressive, and our knowledge is getting more accurate with each passing day, is self-evident. Suffice to compare what we have now with what we had 20 years ago, and its clear that we are not regressing.
Now, as to the specific examples that are falsely used to proove how wrong this wicked science can get:
1. "one hundred published psychology experiments".
Seriously, anyone in the field knows that these studies are rubbish, and all correlations are obtained after running so many statstical tests that they are clearly due to chance.
2. "half of all academic biomedical research will ultimately prove false".
Again, this is obvious. Many initially positive findings will be due to chance, especially when we are talking of small, exploration studies. Yes, many initially promising molecules did not live out to the hype, but others did, and spectacularly so. For instance, progress in oncology has been amazing lately. People who had life expectancies of mere weeks in early 2000s now routinely survive for years with their metastatic cancers.
3. "but are unlikely to mention a similar experiment conducted on reviewers of the prestigious British Medical Journal." [as opposed to Sokal's Social Text hoax].
Apples and oranges. Sokal's article is plain old gibberish, whereas the experiments in medical litterature were to assess the reviewers capacity to assess for methodology failures of otherwise plausible studies. Most of these methodology failures are weeded out by dedicated statistical and clerical staff.
Indeed, I would argue that some of those methodological problems were pretty minor, e.g. :
"Poor justification for conducting the study; No ethics committee approval; Failure to spot word reversal in text leading to wrong interpretation of results; No explanations for ineligible or non-randomized cases; Safety. No mention was made of monitoring patients for untoward effects; Format. The abstract was not written in the structured form requested by Annals; References. No reference cited was more current than 1989 despite the fact that there were numerous more recent studies; 5.Presentation. There were multiple grammatical and spelling mistakes, including the misspelling of propranolol as “propanalol." [2,3]
[1] http://hermiene.net/essays-trans/relativity_of_wrong.html
[2] http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2586872/
[3] http://www.annemergmed.com/article/S0196-0644(98)70006-X/ful...