News

1 October 2022: We welcome Yanfang Xia, who moved from Zurich to Bonn, and Huaiyu Liu as a postdoctoral researcher to our team at UCL, who will work in the ESRC-funded project “A cognitive-computational model of avoidance learning”. After wrapping up the initial phase of the CRPP “Synapse & Trauma” (LINK), our affiliation with University of Zurich has now ended.  

Speakers at the symposium, from left to right: Jason Kerr (Max Planck Institute for Neurobiology of Behavior), Peter Dayan (Max Planck Institute for Biological Cybernetics), Dominik Bach (University of Bonn), Benjamin Grewe (ETH Zurich), Vanessa Stempel (Max Planck Institute for Brain Research), and Bianca A. Silva (Consiglio Nazionale di Ricerche). Photograph courtesy of Meike Brömer/University of Bonn.

25 September 2022: Today was the day of Professor Dominik Bach’s Inaugural Symposium “Circuits for Survival”. If you took the opportunity to attend the lectures, we hope you enjoyed the program. The report is now available online on the website of the Transdisciplinary Research Area “Life and Health”. Read more about the symposium here.

24 September 2022: Team members of all three research locations – Bonn, London, and Zurich – got together in Bonn. In anticipation of the symposium, and a workshop on the following day, the teams met for the first time for dinner on Sunday night.

Photograph courtesy of Lukas Kornemann. 

15 September 2022: Today, we welcome Lukas Kornemann as a doctoral fellow to our team in Bonn. He will work on human escape response in the context of the ERC project “ActionContraThreat”.

06 September 2022: We are delighted to announce, the inaugural lecture of Prof. Dr. Dr. Dominik Bach as the Hertz Chair for Artificial Intelligence and Neuroscience at University of Bonn on 26 September 2022. The lecture will be part of “Circuits for Survival”, a symposium with renowned speakers organized by the Transdisciplinary Research Area “Life and Health” of Bonn University.

20 August 2022: New article published – Survival in biological environments requires learning associations between predictive sensory cues and threatening outcomes. Such aversive learning may be implemented through reinforcement learning algorithms that are driven by the signed difference between expected and encountered outcomes, termed prediction errors (PEs). While PE-based learning is well established for reward learning, the role of putative PE signals in aversive learning is less clear. Here, we used functional magnetic resonance imaging in humans to investigate the neural representation of PEs during maintenance of learned aversive associations. Surprisingly, we find clear PE representations for US omission but not for US presence. More details in the article (open access!). Ojala et al. (2022)

Whole-brain PE fMRI results. Figure from the article on asymmetric representation of aversive prediction errors in Pavlovian threat conditioning linked above.

15 August 2022: Today, we welcome Yonatan Hutabarat as postdoctoral fellow to our team in Bonn. He will work on the structure of threat-related movements in the context of the ERC project “ActionContraThreat”.

01 August 2022: Today Federico Mancinelli joined us at University of Bonn as postdoctoral fellow, where he will work on cognitive-computational models of threat avoidance in the context of the CogLearn project. We also welcomed Merle Ingenfeld as research coordinator to the team.

17 July 2022: We welcome Sascha Bonneschranz as lab manager to our team in Bonn. 

16 July 2022: PsPM 6.0.0 released – including developer mode, gaze signal post processing, specific lateral or combined pupil data processing, improved file overwriting behaviour, supporting PPG data loading, various bug fixes and improvements for functions, and more. Thanks to everybody involved! Get the software here: https://github.com/bachlab/PsPM/releases

01 April 2022: Today is the official starting date of the Hertz Chair for Artificial Intelligence and Neuroscience at University of Bonn, where part of the team will be based. University of Bonn covered the new Chair with a very stimulating press release. We warmly welcome our new team members in Bonn: Jules Brochard joined the ActionContraThreat team as a postdoctoral fellow to work on theoretical/computational models of threat avoidance, and Anja Menke is our new administrative assistant.

29 March 2022: PhD student Yanfang Xia’s research on scan path speed during fear conditioning is covered in this very engaging piece by Melinh K. Lay, published by the Psychonomic Society.

24 February 2022: The role of sensory cortices in learning, including classical conditioning, has begun to be established. Rodent studies have found that sensory cortices are crucial for associating complex, naturalistic auditory stimuli with threat but this is less clear for discriminating simple tones. Human neuroimaging studies have found differential neural responses in sensory cortices to conditioned threat vs. safety stimuli but no studies have shown causal evidence on whether sensory cortices are in fact important for threat learning/memory. In a new paper, Karita Ojala with Matthias Staib and collaborators show that non-invasive brain stimulation on primary sensory cortex inhibited threat memory consolidation. They applied continuous theta-burst transcranial magnetic stimulation (cTBS) on individually localized (with fMRI) primary somatosensory cortex (S1). The 40 seconds of cTBS took place immediately prior to conditioning. They used non-painful electric somatosensory stimulation on left hand fingers as conditioned stimuli (CS) and aversive electric stimulation on the foot as unconditioned stimuli (US) ipsilateral to the stimulated S1. Successful conditioning during the initial post-cTBS learning session was measured with skin conductance and pupil size responses, and we used startle eye-blink responses as an index of threat memory retention in a recall test the next day, after overnight consolidation. During initial conditioning, no significant differences appeared between the CS-contralateral and ipsilateral cTBS groups. That is, cTBS on S1 immediately prior to conditioning did not seem to clearly affect the conditioning process itself. However, during recall test under extinction, clear differences between the CS-contralateral and ipsilateral cTBS groups appeared in differential startle amplitude to threat vs. safe CS. This suggests that cTBS on S1 prior to conditioning may have had longer-term effects, potentially on memory consolidation, that manifested as reduced retention the day after cTBS + conditioning, but not during initial learning (however, see article for nuanced discussion!). Therefore, primary sensory cortex may play an important role in threat conditioning also in humans and could be an accessible target region for memory editing with non-invasive brain stimulation (vs. non-targeted, systemic pharmacological and behavioural interventions). More details, including complex vs. simple stimuli, in the article (open access!). (Ojala*, Staib*, et al., Biological Psychiatry, in press).

02 February 2022: We are expanding and moving. Dominik Bach has accepted the position of a Hertz Chair for Artificial Intelligence and Neuroscience at University of Bonn. Lab transition will start in April 2022 while ongoing projects at UCL will continue over the next few years. We are hiring – check out these job openings.

01 February 2022: We welcome our new team members at UCL. Sajjad Zabbah joined the ActionContraThreat team as a postdoctoral fellow to work on cognitive and neural models of threat avoidance. Marina Rodriguez Lopez is our new research assistant in the ESRC-funded CogLearn project.

07 October 2021: First team dinner in London since COVID started. It’s sad to say goodbye to Jack and Samson and we wish them all the best for their careers.

26 July 2021: PsPM 5.1.1 released – including bugfixes for non-linear model, splitting sessions, and skin conductance response processing, the improved user interface for windows, macOS and Linux, and a new basis function template for data vector. Thanks to everybody involved! Get the software here: https://bachlab.github.io/PsPM/software/

18 June 2021: PsPM 5.1.0 released – including a new function for skin conductance response processing and quality assessment, the improved scanpath speed model, and the new GUI designed with .mlapp. Changes to functions for better warning management and consistent data processing, together with bugfixes, are also included in this release. Thanks to everybody involved!

04 February 2021: All known anxiolytics reduce the frequency of theta oscillations in the hippocampus. Could that explain their anxiolytic action? In a new biophysical model, we demonstrate that even slight theta frequency reduction (e.g. 0.5 Hz) can lead to a dramatic impairment in synaptic memory, including contextual fear memory. The model also explains the known phenomenon of increased amygdala theta power after contextual fear conditioning. (Castegnetti, Bush, Bach, 2021, Hippocampus)

16 November 2020: Many experimental psychologists are interested in latent attributes: memory, attention, confidence, etc. They collect observables, and pre-process/transform them, to measure the latent attribute. For a given attribute, what is the best observation and transformation method? In a new paper in Nature Human Behaviour, we propose calibration experiments, in which one manipulates the latent attribute with an established experimental treatment, and then select observation/transformation/pre-processing method to the maximise correlation between measured latent attribute, and the independent variable (“retrodictive validity”). The calibration treatment does not have to match planned research, and its effect can be weak. We statistically show that the optimised measure will maximise accuracy to infer the latent attribute, independent of the experimental paradigm. Explainer videos: why is calibration useful, how does it work, what is the statistics background. See also our blog post. (Bach, Melinscak, Fleming, Voelkle 2020, Nature Human Behaviour)

10 November 2020: Threat-related cues have been demonstrated to bias covert and overt attention. In a newly published study, we exploited a summary statistic of gaze patterns during CS presentation as an index of human threat conditioning. We observed shorter saccadic scanpath length during CS+ compared to CS-. The potential underlying mechanism appeared to be more and longer fixation on the screen centre during CS+. The effect size of scanpath length is comparable to that of skin conductance responses. Streamlined quantification of scanpath length is implemented in the newest release of PsPM. (Xia, Melinscak, Bach 2020, Behavior Research Methods)

05 November 2020: We welcome new team members who joined us this autumn. Juliana Sporrer is a PhD student in Computational Psychiatry at UCL. Ulises Serratos joined the UCL team as postdoctoral fellow to investigate motion trajectories under threat. Dadi Zhao is our new PsPM programmer. Congratulations to Sarrina Tursunova who passed her MSc exams at UZH today!

08 September 2020: We released PsPM 5.0.0 with several new features such as new options in SCR quality correction, the possibility to convert gaze distances to scan path speed or the possibility to import data from DSV (delimiter separated values) file format. This release also includes several technical changes (e.g. in pupil import/pre-processing) as well as bugfixes. Thanks to everybody involved!

31 August 2020: Anterior hippocampus is essential for balancing threat and reward. Several neuroimaging studies have shown that neural activity in this region relates to threat probability under approach-avoidance conflict. However, a recent lesion study (Bach et al. 2019) indicated hippocampus may be relevant for the decision to avoid rather than for representing threat features. In a new study using high-resolution fMRI, we disentangle representation of threat features (probability and magnitude) and behaviour. While anterior hippocampus BOLD signals relates to threat features, it is better explained by ensuing avoidance decision. This relation to avoidance behaviour was specific to subfields CA2/3, compared to CA1 (Abivardi, Khemka & Bach, 2020, Journal of Neuroscience).

18 July 2020: We welcome new team members who joined us during the COVID-19 pandemic. Samson Hall is supporting our VR development as 3D artist. Samuel Maxwell is our new PsPM programmer, based at UCL.

15 May 2020: Making decisions on probabilistic outcomes requires retrieval and integration of the possible outcomes. In a new study using magnetoencephalography, we let participants play a risky decision-making game and derived magnetic field patterns that discriminated the outcomes. We then decoded outcome-related brain activity during decision making and found sequential (alternating) outcome retrieval that was dependent on loss magnitude and predicted upcoming behaviour (Castegnetti et al., 2020, Nature Communications).

11 May 2020: Taking risks to obtain reward is important, and adolescence is the time to learn this balance. In a new paper, we looked at predictors of risky foraging in N = 781 people aged 14-24. We split the data into a discovery and hold-out sample, and pre-registered all hypotheses before accessing the holdout. Sex is the best predictor of performance and behaviour, with males earning 20% more than females, but not “surviving” less often in this computer game. Further independent (but smaller) impact of IQ, as well as self-reported cognitive complexity and daringness. No effect of anxiety, and surprisingly, of age and maturation. Notably, self-reported anxiety and daringness are uncorrelated in this sample. What may underly the large impact of sex? Our results are in keeping with a recent field study in Uber drivers that found males take bigger risk and consequently earn more. Certain features of our game and of the gig economy (e.g. no cooperation possible, single-shot payment) may disadvantage females. (Bach, Moutoussis et al. 2020, Nature Human Behaviour).

01 April 2020: Just in time for the upcoming workshop, PsPM version 4.3.0 is released with bugfixes and new features. Thanks to everybody involved, in particular Ivan Rojkov! Get the software here.

30 March 2020 COVID-19 update: All research involving participant visits has been paused at both sites due to the pandemic. All meetings and teachings have moved online. Planned PsPM courses at ESCAN and EMHFC have been cancelled by the organisers. Instead, we are offering a free live-online PsPM workshop from 06 April to 20 May 2020. More details on the PsPM website.

16 March 2020: A reminder/extinction procedure has been suggested as a protocol for editing unwanted aversive memories, but evidence from human replication studies is mixed. Most studies have used threat-conditioned SCR as a model system; those that used fear-potentiated startle found no effect of the manipulation. Here, former MSc student Josua Zimmermann measured threat-conditioned pupil size responses (PSR) and SCR, in N=71 participants, and finds no effect of the manipulation in either measure (Zimmermann & Bach 2020, Learning & Memory).

28 February 2020: We are running the next full PsPM workshop on 30.6.-01.07.2020 at ESCAN 2020 in Budapest. This 2-day workshop will include theoretical and practical sessions on eyetracking (pupillometry and gaze analysis), SCR, ECG, respiration and startle-eyeblink measurement. Training will be adapted to participant’s previous exposure to PsPM, SPM or Matlab. Participants are welcome to bring along specific questions on PsPM, or experiments that they are planning to analyse with PsPM.

24 February 2020: Psychophysiological Modelling (PsPM) is being developed with an eye on measurement of fear conditioning. In a new review, we summarise the background and requirements for measuring human fear conditioning, cover technical details of PsPM models used for this purpose, and what PsPM adds. It turns out that required sample sizes to achieve a given level of statistical power can differ up to a factor of three between different observables, and between PsPM and alternative methods (Bach & Melinscak 2020, Behaviour Research and Therapy).

24 January 2020: We welcome Jack Brookes who joined the UCL-based lab as a post doc and will be working on virtual reality serious games. Congratulations to Karita Ojala who successfully defended her PhD thesis at UZH this week!

17 January 2020: Using intuition to design informative experiments, which allow accurate statistical inferences, is challenging. This is especially true in fields where there has been a proliferation of increasingly complex theoretical models. The study of associative learning provides an example of such a field (e.g., see Tzovara et al. 2018). In a new paper, we propose for associative learning experiments to be algorithmically optimized. Through simulations we show that optimized experiments may provide substantially improved statistical inferences. Due to the generality of the simulation-based experimental design framework, the proposed approach may also be useful in other learning and decision making experimental domains (Melinscak & Bach 2020, PLOS Computational Biology).

12 December 2019: Christmas dinner in Zurich – this year enjoying Chinese food while commencing to wind down after a year’s hard work. We also welcomed our new study coordinator Aneta and administrative assistant Ilona who started with us in November.

4 November 2019: We migrated PsPM to GitHub and released version 4.2.1 – including various changes to increase ease of use, several bugfixes and new tests . Thanks to everybody involved, in particular Eshref Yozdemir! Get the software here.

09 October 2017: Classically, amygdala and thalamus are required to learn threat predictions, without a necessary role for cortex. However, some rodent research suggests that to discriminate threat-predicting cues from cues predicting safety, sensory cortex is required. We have recently shown that Heschl’s gyrus in human auditory cortex represents threat associations, over and above the physical difference between threat- and safety-predicting cues, both for physically simple (monophone) and complex (spatiotemporal patterns) sounds. Here, we replicate this finding for a different class of sounds without temporal discriminative features, and confirm it is restricted to primary auditory cortex within Heschl’s gyrus (Staib et al. 2019, Human Brain Mapping, in press). In ongoing work, we seek to establish whether these representations are causally required for threat learning. 

17 September 2019: Last Wednesday, September 11th, our traditional Summer BBQ took place – this time together with our three partner labs (CRPP Synapse & Trauma). We spent a cheerful evening with colleagues and friends at the nearby Quartierhof Wynegg. Ahead of the official part of the evening we said goodbye to our PhD Student, Karita Ojala, who sadly leaves us after three years’ time as well as to Lena Rhonheimer, our Student Research Assistant for the past year. Thank you both for your commitment – it was a pleasure having you in our team! We wish you all the best for the future and a lot of success with your endeavours!

16 September 2019: PsPM 4.2.0 released – including new pupil preprocessing, pupil foreshortening error correction and QRS detection algorithms; performance optimizations to pupil data import functions; new tests, bugfixes and changes to the behaviour of some functions to make PsPM more uniform and easier to use. Thanks to everybody involved! Get the software here.

12 September 2019: MMP-9 inhibition appears to attenuate memory consolidation (see our previous paper: (Bach et al 2018, Molecular Psychiatry). It could also be a target for blocking reconsolidation. In a new paper, we tested this hypothesis in human threat conditioning. We find that doxycycline has no specific impact on a reminded cue, but confers a global reduction in extinction learning and threat learning beyond the clearance of the drug. This may point towards a more long-lasting impact of doxycycline treatment on memory plasticity (Bach et al. 2019b, Journal of Neuroscience).

02 September 2019: Approach-avoidance conflict behaviour requires anterior hippocampus, and in a range of tests also amygdala. Most of these tests require a number of individual actions, and which of these rely on hippocampus or amygdala is unclear. In a clinical lesion study together with our colleagues at the Charité Department of Neurology in Berlin, we show that human hippocampus lesions impair the decision to stay away from reward when danger is high, but leave swift escape from threat intact. Amygdala lesions, in contrast, do not impair the decision to stay away from reward but markedly increase the vigour of escape from threat (Bach et al. 2019a, Journal of Neuroscience).

14 August 2019: Dominik Bach has accepted an appointment as Principal Research Fellow at the UCL Queen Square Institute of Neurology, University College London. He will be joining the Max-Planck/UCL Centre for Computational Psychiatry and Ageing Research, and the Wellcome Centre for Human Neuroimaging, as a Principal Investigator later in the year, while research activities continue in Zurich.

11 July 2019: Many ethological approach-avoidance conflict tests do not allow separating different action components. This impedes investigation whether they are under the same or dissociable neural control. We have previously developed a human computer game to separate action components (Bach 2015 PLOS CB). Here, our collaborators at Platrad build on this idea and present a novel 2-lever approach-avoidance conflict test for mice that separates decision to approach, approach latency, and approach vigour (Oberrauch et al., 2019, Psychopharmacology).

04 July 2019: Magnetoencephalography (MEG) measures the magnetic field emitted by the brain and allows reconstructing the underlying neural activity. For deep sources such as the amygdala, it has been unclear whether what appears as source activity is really coming from this source, or from nearby areas. In this new paper, Athina Tzovara and colleagues at UZH and UCL record high-precision MEG. They then show that the measured data can be better explained with reconstruction models that allow for sources in the amygdala than for models that do not. Furthermore, just moving this anatomical location by 3 mm made the explained variance drop decisively. This demonstrates the spatial precision of the method. They then analyse amygdala source data during threat conditioning and find a pattern that is rather different from what is commonly observed in short rodent experiments (Tzovara et al., 2019 Human Brain Mapping).

17 June 2019: PsPM 4.1.1 released – including new models and data processing methods, support for two new eye tracker formats, various improvements and bugfixes. Thanks to everybody involved! Get the software here.

4 June 2019: In this new article, the CRPP PIs Dominik Bach, Steven Brown, Birgit Kleim and Shiva Tyagarajan, review evidence for roles of the extracellular matrix in learning and memory, and in experience-dependent psychiatric conditions. They then suggest how control of the extracellular matrix could be leveraged for innovative treatments and discuss possible aetiological mechanisms of extracellular matrix alterations in psychiatric disorders. Swiss Medical Weekly, 149:w20060

2 May 2019: In a new paper, Christoph Korn investigates how humans decide to forage under combined risk of starvation and predation in an virtual approach-avoidance conflict task (Korn & Bach, 2019, Nature Human Behaviour). He shows that participants rely both on predator probability, and on the normatively optimal policy. Predator probability was related to BOLD signals in hippocampus and amygdala. The optimal policy was related to medial prefrontal cortex.