Field: Neuroscience › Sensory SystemsType: No effect foundArea: tinnitusTarget: Game-based frequency discrimination trainingStage: randomized trial
Hypothesis
Building auditory frequency training into a computer game would make people more motivated, so they would train more and their tinnitus would improve more than with standard task-based training.
What failed
In 60 participants over six weeks, the game versions felt more motivating, but compliance was about 70% in all three groups and tinnitus severity and attention did not change differently between groups.
Broken assumption
Making a therapy feel more enjoyable increases how much people do it and how much it helps.
Why
Higher reported motivation did not lead to more training or to more benefit.
Field: Psychology › Social PsychologyType: Couldn't replicate a published resultArea: parapsychologyTarget: Retroactive facilitation of recall (Bem)Stage: replication
Hypothesis
Practising words after a memory test would improve recall of those words in the test, as one of Bem's precognition experiments reported.
What failed
Three pre-registered exact replications found no significant effect (combined n = 150, combined p = .83, one-tailed).
Broken assumption
Events after a test can influence performance on the test.
Field: Neuroscience › Cellular and Molecular NeuroscienceType: No effect foundArea: axon guidance signallingTarget: EphrinA5 binding to RET receptor tyrosine kinaseStage: preclinical
Hypothesis
EphrinA5 would bind the RET receptor directly, as earlier cell-based assays suggested, which would explain reverse signalling that guides motor axons.
What failed
Zebrafish ephrinA5b bound mouse EphA4 with high affinity, but in vitro it showed no interaction with zebrafish RET.
Broken assumption
EphrinA5 and RET bind each other directly.
Why
The authors suggest that other molecules may be needed for ephrinA5 to signal through RET, which cell-based assays would contain and a purified system would not.
Field: Biochemistry, Genetics and Molecular Biology › Cell BiologyType: A method hit a wallArea: skin tissue engineeringTarget: Serum-free, feeder-free keratinocyte cultureStage: preclinical
Hypothesis
Skin cells (keratinocytes) grown in serum-free, feeder-free conditions, which avoid animal contaminants, would still form layered skin in a reconstructed skin model.
What failed
Keratinocytes grown only in serum-free, feeder-free conditions did not form a layered, mature epidermis, and adding calcium did not help.
Broken assumption
Culture conditions that support skin cell growth in a flat dish also support building three-dimensional skin.
Why
The cells had not lost their potential: they formed layered skin when serum was added back. Serum seems to contain factors that both help and block layering.
Field: Neuroscience › Cognitive NeuroscienceType: Couldn't replicate a published resultArea: computational neuroscienceTarget: Chaotic rate dynamics in spiking neuron networksStage: replication
Hypothesis
Networks of spiking neurons would show the same chaotic instability as matched rate networks, as an earlier study proposed.
What failed
The rate networks showed the expected signs of chaos, such as autocorrelation times that diverge near the transition. In the spiking networks, the autocorrelation time did not depend on coupling strength.
Broken assumption
Spiking neuron networks become chaotic in the same way as rate-based network models.
Why
The mathematical analogy between rate networks and spiking networks did not hold in the spiking simulations.
Field: Medicine › OncologyType: A method hit a wallArea: inflammatory disease immunotherapyTarget: Neutralising mouse antibodies against human CCR6Stage: preclinical
Hypothesis
Immunising mice with CCR6 peptides or CCR6-expressing cells would produce monoclonal antibodies that block the receptor from binding its ligand CCL20.
What failed
Peptide immunisation produced no antibodies that recognised the native receptor. Cell immunisation produced CCR6-specific antibodies, but none blocked CCL20 binding.
Broken assumption
Standard mouse immunisation can produce function-blocking antibodies against human CCR6.
Why
All the antibodies, including commercial ones, bound one dominant region at the start of the receptor that does not affect its function.
Field: Biochemistry, Genetics and Molecular Biology › Molecular BiologyType: Couldn't replicate a published resultArea: pain therapeuticsTarget: SVmab1 antibody against NaV1.7Stage: preclinical
Hypothesis
The monoclonal antibody SVmab1, made from its published sequence, would bind the NaV1.7 sodium channel and block its currents, as originally reported.
What failed
Recombinant SVmab1 did not bind the peptide it was raised against or the purified NaV1.7 domain, did not bind NaV1.7 on three types of live cells, and blocked NaV1.7 currents no more than a control antibody.
Broken assumption
An antibody can be reproduced with the same activity from its published sequence.
Why
The abstract does not give a cause. The antibody made from the published sequence did not have the reported activity.
Field: Environmental Science › Global and Planetary ChangeType: A method hit a wallArea: fisheries scienceTarget: Otolith chemistry of skipjack tuna nursery areasStage: full study
Hypothesis
The chemistry of ear bones (otoliths) in young skipjack tuna would differ enough between the three main Indian Ocean nursery areas to tell where an unknown fish was born.
What failed
Some single elements differed between nurseries, but the overall chemical signatures largely overlapped. The best classifier assigned fish to the correct nursery only 51% of the time, and the Central and Eastern nurseries were hard to separate.
Broken assumption
Otolith chemistry is distinct enough between Indian Ocean nursery areas to identify the birthplace of skipjack tuna.
Why
The nursery signatures overlap, and the signatures also changed between the 2018 and 2019 cohorts, which makes a fixed reference signature unreliable.
Field: Immunology and Microbiology › ImmunologyType: No effect foundArea: hivTarget: Cytosolic DNA sensing (IFI16–STING) in CD4+ T cellsStage: preclinical
Hypothesis
Human T cells, which contain the DNA-sensing machinery, would respond to foreign DNA in the cytoplasm by making type I interferon and restricting HIV.
What failed
The DNA sensor IFI16 moved to foreign DNA and recruited STING and TBK1, but the T cells did not make interferon or interferon-stimulated genes, and cytoplasmic DNA did not affect HIV replication.
Broken assumption
Cells that contain the DNA-sensing proteins can use them to start an interferon response.
Why
The authors propose that the DNA signalling pathway is impaired somewhere after STING and TBK1 in T cells. Added interferon still suppressed HIV.
Field: Neuroscience › Cognitive NeuroscienceType: A method hit a wallArea: neural engineering and somatosensory prostheticsTarget: Direct cortical stimulation of primary somatosensory cortex as prosthetic feedbackStage: full study
Hypothesis
Stimulating somatosensory cortex can stand in for touch in a neuroprosthesis, giving feedback as usable and as fast as a cutaneous haptic cue.
What failed
In four subjects implanted with electrocorticographic grids, median response times to reliably perceived cortical stimulation ran from 254 ms to 528 ms against 198-313 ms for haptic stimuli to the hand, a significant slowing; no learning or habituation appeared across blocks and stimulation train duration did not change the latency.
Broken assumption
Cortical stimulation is perceived as quickly as peripheral touch, so it can be substituted for it in a feedback loop.
Why
The authors offer no mechanism, presenting the latencies as a realistic expectation for prosthetic feedback and calling for study of how cortical stimulation is perceived.