As a dissertation and a psychology study, the small sample size is somewhat expected, and a power analysis would (i suspect) indicate that it is perfectly valid for the statistical test being used (36 subjects is pretty close to 80% power for a moderate effect size for simpler experimental designs like this, if I remember correctly). These studies are not intended to serve as a critical and seminal authority on the subject but to add evidence to the literature, which then is taken as a whole to understand the underlying phenomena. Especially in a world where grants are reserved for only the flashiest and trendiest technologies and topics and publish or perish rules the day, papers with small sample sizes are a symptom of a lack of funding for these questions and relentless demands on researchers' time and effort, and I would bet that these researchers were doing the best they could with the resources they had. I think hackernews is rather uncharitable about psychology research in these respects.
As a note, this study was published in 2009 and thus likely conducted at least a year or two prior, at which point preregistration was not common for the cognitive field (and indeed it is still in its infancy). So, it is not surprising that this study was not preregistered and nor, I would argue, should it be considered a red flag.