# A replication crisis in psychological science Psychological science has been under serious scrutiny, since a few decades now. Academic psychology suffers from a ‘replicability crisis’. This seemed to hit social psychology, and priming studies, in particular (cf. Kahneman 2012), but other psychological fields are not immune to it (Open Science Collaboration 2015). What is the problem? Many psychological experiments have shown to be unable to yield the same results upon repetition. Some of these experiments have resulted in theories that appeared in well-known journals and well-cited publications. Some have become part of psychological theorising for several decades. When a particular finding cannot be repeated, many theorising based on it may be lacking a foundation. How can it happen that a scientific result turns out to have been a one-off result? A psychological theory aims to understand, and ultimately predict, human mental activity and behaviour. As these are probably among the most complex phenomena occurring in the universe as we know it (with the logical exception of the whole universe itself), there is ample opportunity for uncontrolled variables to influence the results of studies into these phenomena. The complexity of human activity arises from a myriad of causes, many escaping control of the experimenter. The behaviour (and mental activity) of a human subject can be caused by current and past behaviour, perceptions, ideas, surrounding and physiology (to name just a few). In order for a psychological theory to be valid and reliable, many aspects of behaviour and mental activity ideally are captured by it. Particularly the many potential causes of, and influences on, the phenomena to be explained should be acknowledged to play a role in the construction of the theory. Overlooking them may result in theories that falter in replication research. For above mentioned reasons I suggest four properties that a psychological theory should have in order to minimise the risk of becoming unreplicable, viz. they should be Holistic, Embodied, Phenomenological and Implicit). ## Holistic A theory should include as many ecologically valid aspects as viable. For theories of sensory perception for example this means the aspects are based on multiple sensory inputs. All sensory systems interact, with each other but also with top down processes like expectations (cf. Ellingsen et al. 2016, Piqueras-Fiszman and Spence 2016). The stimulation employed to study, propose or test a specific theory should be close to the situation that the theory is supposed to work in. This means that a non-holistic (reductionist) approach will increase the chance of a theory faltering in subsequent experiments, where some part of the situation may be different yet (in hindsight) important. The additional effects of top down processes imply that the information that subjects receive –or may infer- becomes part of the experiment, and should ideally be under rigorous control. Although this seems to argue pro strictly isolated (reductionist) laboratory studies –to control as much unwanted influences as possible-, such studies can at the same time often be un-holistic. This dilemma is certainly a dilemma known to applied psychological scientists. They often find themselves working in noisy environments, trying to control the noise. I’m not referring to sound, but to ‘experimental noise, i.e. unaccounted for influences on their data. If they obtain an effect in their study it was large enough to stand out from the noise, and valid (from the point of view of their applied research question) as it appears under those noisy, realistic, circumstances. Studies in, low noise, laboratories may yield significant, but sometimes rather small, effects that can be relevant for testing a specific theory, but outside the laboratory (or in another laboratory) the effect may not appear. A practical way out is performing a ‘proof-of-principle’ or ‘pilot’ study first, followed by larger studies and finally by real-life field studies. ## Embodied A psychological theory should recognise that cognition is hardly an abstract and isolated mental activity but that it is grounded, primarily in the body (cf. Barsalou 1999, 2008). Psychological activity, be it thinking, perceiving, or behaving cannot be seen separate from the interaction of a subject with its body (Semin and Smith 2008, Shapiro 2014). In addition the state this body is in must be taken into account as it can have serious effects on cognition. This can be seen in decision making situations. Look at Bechara et al. (1997), where a relation to Somatic Marker theory (Damasio 1996) may exist (cf. Reimann et al. 2012), but perhaps this should be called embodied emotion rather than cognition. We will not expand on this here. Although a general physiological state of the body (hunger, fear, fatigue, etc.) is not what is commonly meant in ‘embodied cognition’, it can exert an influence on cognition and behaviour. For this reason it is mentioned here. ## Phenomenological A psychological theory should be phenomenological, it should refer to what an individual experiences, instead of to some description of what is presented to the subject. External (physical or semantic, sometimes called ‘objective’) descriptions of stimuli can have an unclear, unknown, relation to the percepts the subject experiences. Percepts could arise as a result of an illusion or of an interaction between information and sensory stimulation. This potentially destroys the relation of an external (‘objective’) description of the stimulation to a subject’s experience. Although the stimulation presented to subjects in a psychological experiment should be described as completely as possible in ‘external’ terms, it should be clear that there can be, and often is, a difference between the presentation of this stimulation and the experience by the subject as a result from that stimulation. ## Implicit A psychological theory should take into account the fact that most psychological processes remain inaccessible to the individual itself. This means that a theory should ideally not rely on (only) self-reports. There is a difference between ‘reasons’ and ‘causes’, just like correlation does not imply causation. The ‘snare of psychology’ (James 1890) refers to the fact that the state that a reporting subject is in, is different from the state the subject is (trying to) report on. The act of thinking about a question, interpreting it and formulating an answer is likely to distort the very thing under investigation. This may in particular affect questionnaire based studies where a theory is based on a Factor Analysis of a list of items. Posing an ‘_n_\-Factor model’ as a theory in a specific field, based on questionnaires alone is likely to yield a theory that may be hard to replicate under (even slightly) different circumstances.