The F1000Research will be waiving the publication fee for negative result manuscripts up through the end of August.
If you have negative results in your lab notebooks, this is the time to write them up! Like all journals, we of course publish traditional full-length research papers but, in addition, we accept short single-observation articles, data articles (i.e. a dataset plus protocol), and negative- and null-result submissions.
For negative and null results, it is especially important to ensure that the outcome is a genuine finding generated by a well executed experiment, and not simply the result of poorly conducted work. We have been talking to our Editorial Board about how to try to avoid the publication of the latter type of result and will be addressing this topic and asking for your input in a further post in the next few days.
The follow up post requesting comment is here.
This is a great idea and the original post nails down why.
This is not only a disappointment for the researchers who conducted the work, it’s also damaging to the overall scientific record. This so-called “publication bias” toward positive results makes it appear as though the experiments with negative or null results never happened.
Sometimes the unpublished experiments are obvious next steps in elucidating a particular biological mechanism, making it likely that other researchers will try the same thing, not realizing that someone else already did the work. This is a waste of time and money.
On other occasions, the positive results that are published are the exception: they could have been specific to a narrow set of conditions, but if all the experiments that didn’t work are not shown, these exceptional cases now look like the only possible result. This is especially damaging when it comes to drug development and medical research, where treatments may be developed based on an incomplete understanding of research results.
The waste of time and money cannot be emphasized enough, especially in these tight funding times. Why on earth should we tolerate any duplication of effort that is made necessary simply by the culture of not publicizing results that are not deemed sexy enough? This is the information age, people!
One example from my field is the self-administration of delta9-tetrahydrocannabinol (THC) by the common laboratory species used for self-administration studies of other drugs of abuse. Papers by Goldberg and colleagues (Tanda et al, 2000; Justinova et al, 2003) showed that squirrel monkeys will self-administer THC intravenously which was big news. It was the first relatively clear demonstration in lab animals for a substance we know humans readily self-administer. As the Goldberg group related in their 2005 review article, there is no clear evidence that rodents will self-administer THC i.v. in literature stretching back to the 1970s when the self-administration technique was being used for studies of numerous drugs.
Over the last three decades, many attempts to demonstrate intravenous self-administration of THC or of synthetic cannabinoid CB1 receptor agonists by experimental animals were relatively unsuccessful (Pickens et al., 1973; Kaymakcalan, 1973; Harris et al., 1974; Carney et al., 1977; van Ree et al., 1978; Mansbach et al., 1994) (Table 1). None of these studies clearly demonstrated persistent, dose-related, self-administration behavior maintained by THC or synthetic cannabinoids, which would be susceptible to vehicle extinction and subsequent reinstatement in the absence of unusual ‘‘foreign’’ conditions.
The thing is that rats "wouldn't" self-administer nicotine either. Nor alcohol. That is, until people came up with the right conditions to create a useful model. In the case of ethanol it was helpful to either force them to become dependent first (via forced liquid diets adulterated with ethanol or ethanol inhalation chambers) or to slowly train them up on cocktails (called the flavorant-fade procedure). In the case of nicotine, the per-infusion dose was all critical and it helped to provide intermittent access, e.g., with four days on, three days off. Interestingly, while making rats dependent on nicotine using subcutaneous osmotic pumps didn't work (as it does for heroin) very well, a recent study suggests that force inhalation-based dependence on nicotine results in robust intravenous self-administration.
For many drugs of abuse, subtle factors can make a difference in the rodent model. Strain, sex, presence of food restriction, exact age of animals, circadian factors, per-infusion dose, route of administration, duration of access, scheduling of access.... the list goes on and on. A fair read of the literature suggests that when you have cocaine or heroin, many factors have only quantitative effects. You can move the means around, even to the p<0.05 level, but hey, it's cocaine or heroin! They'll still exhibit clear evidence that they like the drug.
When it comes to other drugs, maybe it is a little trickier. The balance between pleasurable and aversive effects may be a fine one (ever tried buccal nicotine delivery via chew or dip? huh?). The route of administration may be much more critical. Etc.
So the curious person might ask, how much has been tried? How many curious grad students or even postdocs have "just tried it" for a few months or a year? How many have done the most obvious manipulations and failed? How many have been told to give it up as a bad lot by older and wiser PIs (who tried to get THC self-administration going themselves back 20 years ago)?
I'm here to tell you that it has been attempted a lot more than has been published. Because the lab lore type of advice keeps rolling.
It is really hard, however, to get a comprehensive look at what has been tried and has led to failure. What were the quality of those attempts? N=8 and out? Or did some poor sucker run multiple groups with different infusion doses? Across the past thirty years, how many of the obvious tweaks have been unsuccessful?
Who cares, right? Well, my read is that there are some questions that keep coming around, sometimes with increased urgency. The current era of medical marijuana legalization and tip-toeing into full legalization means that we're under some additional pressure to have scientific models. The explosion of full-agonist cannabimimetic products (K2, Spice, Spike, etc containing JWH-018 at first and now a diversity of compounds) likewise rekindles interest. Proposals that higher-THC marijuana strains increase dependence and abuse could stand some controlled testing....if we only had better models.
Well, this is but one example. I have others from the subfields of science that are of my closest interests. I think it likely that you, Dear Reader, if you are a scientist can come up with examples from your own fields where the ready availability of all the failed studies would be useful.