One might suggest just going to the gym or doing other activities which may increase endogenous testosterone secretion. This would be unsatisfying to me as it introduces confounds: the exercise may be doing all the work in any observed effect, and certainly can’t be blinded. And blinding is especially important because the 2011 review discusses how some studies report that the famed influence of testosterone on aggression (eg. Wedrifid’s anecdote above) is a placebo effect caused by the folk wisdom that testosterone causes aggression & rage!
This doesn’t fit the U-curve so well: while 60mg is substantially negative as one would extrapolate from 30mg being ~0, 48mg is actually better than 15mg. But we bought the estimates of 48mg/60mg at a steep price - we ignore the influence of magnesium which we know influences the data a great deal. And the higher doses were added towards the end, so may be influenced by the magnesium starting/stopping. Another fix for the missingness is to impute the missing data. In this case, we might argue that the placebo days of the magnesium experiment were identical to taking no magnesium at all and so we can classify each NA as a placebo day, and rerun the desired analysis:
If you haven’t seen the movie, imagine unfathomable brain power in capsule form. Picture a drug from another universe. It can transform an unsuccessful couch potato into a millionaire financial mogul. Ingesting the powerful smart pill boosts intelligence and turns you into a prodigy. Its results are instant. Sounds great, right? If only it were real.
Most people would describe school as a place where they go to learn, so learning is an especially relevant cognitive process for students to enhance. Even outside of school, however, learning plays a role in most activities, and the ability to enhance the retention of information would be of value in many different occupational and recreational contexts.
SOURCES: Marvin Hausman, MD, CEO, Axonyx Inc. Axel Unterbeck, PhD, president, chief scientific officer, Memory Pharmaceuticals. Martha Farah, PhD, professor, department of psychiatry, University of Pennsylvania. Howard Gardner, PhD, Hobbs Professor of Education and Cognition, Harvard Graduate School of Education. Nature Reviews Neuroscience, May 2004. Neurology, July 2002. Alzheimer's Association.

This doesn’t fit the U-curve so well: while 60mg is substantially negative as one would extrapolate from 30mg being ~0, 48mg is actually better than 15mg. But we bought the estimates of 48mg/60mg at a steep price - we ignore the influence of magnesium which we know influences the data a great deal. And the higher doses were added towards the end, so may be influenced by the magnesium starting/stopping. Another fix for the missingness is to impute the missing data. In this case, we might argue that the placebo days of the magnesium experiment were identical to taking no magnesium at all and so we can classify each NA as a placebo day, and rerun the desired analysis:
Taking the tryptophan is fairly difficult. The powder as supplied by Bulk Nutrition is extraordinarily dry and fine; it seems to be positively hydrophobic. The first time I tried to swallow a teaspoon, I nearly coughed it out - the power had seemed to explode in my mouth and go down my lungs. Thenceforth I made sure to have a mouth of water first. After a while, I took a different tack: I mixed in as much Hericium as would fit in the container. The mushroom powder is wetter and chunkier than the tryptophan, and seems to reduce the problem. Combining the mix with chunks of melatonin inside a pill works even better.
Enhanced learning was also observed in two studies that involved multiple repeated encoding opportunities. Camp-Bruno and Herting (1994) found MPH enhanced summed recall in the Buschke Selective Reminding Test (Buschke, 1973; Buschke & Fuld, 1974) when 1-hr and 2-hr delays were combined, although individually only the 2-hr delay approached significance. Likewise, de Wit, Enggasser, and Richards (2002) found no effect of d-AMP on the Hopkins Verbal Learning Test (Brandt, 1991) after a 25-min delay. Willett (1962) tested rote learning of nonsense syllables with repeated presentations, and his results indicate that d-AMP decreased the number of trials needed to reach criterion.
If you want to make sure that whatever you’re taking is safe, search for nootropics that have been backed by clinical trials and that have been around long enough for any potential warning signs about that specific nootropic to begin surfacing. There are supplements and nootropics that have been tested in a clinical setting, so there are options out there.
Potassium citrate powder is neither expensive nor cheap: I purchased 453g for $21. The powder is crystalline white, dissolves instantly in water, and largely tasteless (sort of saline & slightly unpleasant). The powder is 37% potassium by weight (the formula is C6H5K3O7) so 453g is actually 167g of potassium, so 80-160 days’ worth depending on dose.
2 commenters point out that my possible lack of result is due to my mistaken assumption that if nicotine is absorbable through skin, mouth, and lungs it ought to be perfectly fine to absorb it through my stomach by drinking it (rather than vaporizing it and breathing it with an e-cigarette machine) - it’s apparently known that absorption differs in the stomach.
But there are some potential side effects, including headaches, anxiety and insomnia. Part of the way modafinil works is by shifting the brain’s levels of norepinephrine, dopamine, serotonin and other neurotransmitters; it’s not clear what effects these shifts may have on a person’s health in the long run, and some research on young people who use modafinil has found changes in brain plasticity that are associated with poorer cognitive function.

The stop-signal task has been used in a number of laboratories to study the effects of stimulants on cognitive control. In this task, subjects are instructed to respond as quickly as possible by button press to target stimuli except on certain trials, when the target is followed by a stop signal. On those trials, they must try to avoid responding. The stop signal can follow the target stimulus almost immediately, in which case it is fairly easy for subjects to cancel their response, or it can come later, in which case subjects may fail to inhibit their response. The main dependent measure for stop-signal task performance is the stop time, which is the average go reaction time minus the interval between the target and stop signal at which subjects inhibit 50% of their responses. De Wit and colleagues have published two studies of the effects of d-AMP on this task. De Wit, Crean, and Richards (2000) reported no significant effect of the drug on stop time for their subjects overall but a significant effect on the half of the subjects who were slowest in stopping on the baseline trials. De Wit et al. (2002) found an overall improvement in stop time in addition to replicating their earlier finding that this was primarily the result of enhancement for the subjects who were initially the slowest stoppers. In contrast, Filmore, Kelly, and Martin (2005) used a different measure of cognitive control in this task, simply the number of failures to stop, and reported no effects of d-AMP.
In paired-associates learning, subjects are presented with pairs of stimuli and must learn to recall the second item of the pair when presented with the first. For these tasks, as with tasks involving memory for individual items, there is a trend for stimulants to enhance performance with longer delays. For immediate measures of learning, no effects of d-AMP or MPH were observed by Brumaghim and Klorman (1998); Fleming et al. (1995); Hurst, Radlow, and Weidner (1968); or Strauss et al. (1984). However, when Hurst et al.’s subjects were tested a week later, they recalled more if their initial learning had been carried out with d-AMP than with placebo. Weitzner (1965) assessed paired-associates learning with an immediate cued-recall test and found facilitation when the associate word was semantically related to the cue, provided it was not also related to other cue words. Finally, Burns, House, French, and Miller (1967) found a borderline-significant impairment of performance with d-AMP on a nonverbal associative learning task.
“My husband and I (Ryan Cedermark) are so impressed with the research Cavin did when writing this book. If you, a family member or friend has suffered a TBI, concussion or are just looking to be nicer to your brain, then we highly recommend this book! Your brain is only as good as the body’s internal environment and Cavin has done an amazing job on providing the information needed to obtain such!”
The research literature, while copious, is messy and varied: methodologies and devices vary substantially, sample sizes are tiny, the study designs vary from paper to paper, metrics are sometimes comically limited (one study measured speed of finishing a RAPM IQ test but not scores), blinding is rare and unclear how successful, etc. Relevant papers include Chung et al 2012, Rojas & Gonzalez-Lima 2013, & Gonzalez-Lima & Barrett 2014. Another Longecity user ran a self-experiment, with some design advice from me, where he performed a few cognitive tests over several periods of LLLT usage (the blocks turned out to be ABBA), using his father and towels to try to blind himself as to condition. I analyzed his data, and his scores did seem to improve, but his scores improved so much in the last part of the self-experiment I found myself dubious as to what was going on - possibly a failure of randomness given too few blocks and an temporal exogenous factor in the last quarter which was responsible for the improvement.

This would be a very time-consuming experiment. Any attempt to combine this with other experiments by ANOVA would probably push the end-date out by months, and one would start to be seriously concerned that changes caused by aging or environmental factors would contaminate the results. A 5-year experiment with 7-month intervals will probably eat up 5+ hours to prepare <12,000 pills (active & placebo); each switch and test of mental functioning will probably eat up another hour for 32 hours. (And what test maintains validity with no practice effects over 5 years? Dual n-back would be unusable because of improvements to WM over that period.) Add in an hour for analysis & writeup, that suggests >38 hours of work, and 38 \times 7.25 = 275.5. 12,000 pills is roughly $12.80 per thousand or $154; 120 potassium iodide pills is ~$9, so \frac{365.25}{120} \times 9 \times 5 = 137.

Another classic approach to the assessment of working memory is the span task, in which a series of items is presented to the subject for repetition, transcription, or recognition. The longest series that can be reproduced accurately is called the forward span and is a measure of working memory capacity. The ability to reproduce the series in reverse order is tested in backward span tasks and is a more stringent test of working memory capacity and perhaps other working memory functions as well. The digit span task from the Wechsler (1981) IQ test was used in four studies of stimulant effects on working memory. One study showed that d-AMP increased digit span (de Wit et al., 2002), and three found no effects of d-AMP or MPH (Oken, Kishiyama, & Salinsky, 1995; Schmedtje, Oman, Letz, & Baker, 1988; Silber, Croft, Papafotiou, & Stough, 2006). A spatial span task, in which subjects must retain and reproduce the order in which boxes in a scattered spatial arrangement change color, was used by Elliott et al. (1997) to assess the effects of MPH on working memory. For subjects in the group receiving placebo first, MPH increased spatial span. However, for the subjects who received MPH first, there was a nonsignificant opposite trend. The group difference in drug effect is not easily explained. The authors noted that the subjects in the first group performed at an overall lower level, and so, this may be another manifestation of the trend for a larger enhancement effect for less able subjects.
Somewhat ironically given the stereotypes, while I was in college I dabbled very little in nootropics, sticking to melatonin and tea. Since then I have come to find nootropics useful, and intellectually interesting: they shed light on issues in philosophy of biology & evolution, argue against naive psychological dualism and for materialism, offer cases in point on the history of technology & civilization or recent psychology theories about addiction & willpower, challenge our understanding of the validity of statistics and psychology - where they don’t offer nifty little problems in statistics and economics themselves, and are excellent fodder for the young Quantified Self movement4; modafinil itself demonstrates the little-known fact that sleep has no accepted evolutionary explanation. (The hard drugs also have more ramifications than one might expect: how can one understand the history of Southeast Asia and the Vietnamese War without reference to heroin, or more contemporaneously, how can one understand the lasting appeal of the Taliban in Afghanistan and the unpopularity & corruption of the central government without reference to the Taliban’s frequent anti-drug campaigns or the drug-funded warlords of the Northern Alliance?)
Another study on the olfactory impact of essential oils like lavender and rosemary revealed that they produced positive effects on cognitive performance and mood.16 And in another study on the inhalation of essential oils like grapefruit, fennel, Estragon, and black pepper essential oil, inhalation of the oils resulted in modulation of sympathetic activity in adults.17,18
The fish oil can be considered a free sunk cost: I would take it in the absence of an experiment. The empty pill capsules could be used for something else, so we’ll put the 500 at $5. Filling 500 capsules with fish and olive oil will be messy and take an hour. Taking them regularly can be added to my habitual morning routine for vitamin D and the lithium experiment, so that is close to free but we’ll call it an hour over the 250 days. Recording mood/productivity is also free a sunk cost as it’s necessary for the other experiments; but recording dual n-back scores is more expensive: each round is ~2 minutes and one wants >=5, so each block will cost >10 minutes, so 18 tests will be >180 minutes or >3 hours. So >5 hours. Total: 5 + (>5 \times 7.25) = >41.

The abuse liability of caffeine has been evaluated.147,148 Tolerance development to the subjective effects of caffeine was shown in a study in which caffeine was administered at 300 mg twice each day for 18 days.148 Tolerance to the daytime alerting effects of caffeine, as measured by the MSLT, was shown over 2 days on which 250 g of caffeine was given twice each day48 and to the sleep-disruptive effects (but not REM percentage) over 7 days of 400 mg of caffeine given 3 times each day.7 In humans, placebo-controlled caffeine-discontinuation studies have shown physical dependence on caffeine, as evidenced by a withdrawal syndrome.147 The most frequently observed withdrawal symptom is headache, but daytime sleepiness and fatigue are also often reported. The withdrawal-syndrome severity is a function of the dose and duration of prior caffeine use…At higher doses, negative effects such as dysphoria, anxiety, and nervousness are experienced. The subjective-effect profile of caffeine is similar to that of amphetamine,147 with the exception that dysphoria/anxiety is more likely to occur with higher caffeine doses than with higher amphetamine doses. Caffeine can be discriminated from placebo by the majority of participants, and correct caffeine identification increases with dose.147 Caffeine is self-administered by about 50% of normal subjects who report moderate to heavy caffeine use. In post-hoc analyses of the subjective effects reported by caffeine choosers versus nonchoosers, the choosers report positive effects and the nonchoosers report negative effects. Interestingly, choosers also report negative effects such as headache and fatigue with placebo, and this suggests that caffeine-withdrawal syndrome, secondary to placebo choice, contributes to the likelihood of caffeine self-administration. This implies that physical dependence potentiates behavioral dependence to caffeine.
With this experiment, I broke from the previous methodology, taking the remaining and final half Nuvigil at midnight. I am behind on work and could use a full night to catch up. By 8 AM, I am as usual impressed by the Nuvigil - with Modalert or something, I generally start to feel down by mid-morning, but with Nuvigil, I feel pretty much as I did at 1 AM. Sleep: 9:51/9:15/8:27
When I worked on the Bulletproof Diet book, I wanted to verify that the effects I was getting from Bulletproof Coffee were not coming from modafinil, so I stopped using it and measured my cognitive performance while I was off of it. What I found was that on Bulletproof Coffee and the Bulletproof Diet, my mental performance was almost identical to my performance on modafinil. I still travel with modafinil, and I’ll take it on occasion, but while living a Bulletproof lifestyle I rarely feel the need.