Edited by: Wendy Wood, University of Southern California, United States
Reviewed by: Jan De Houwer, Ghent University, Belgium; Blair T. Johnson, University of Connecticut, United States
This article was submitted to Cognitive Science, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Habits are a powerful route to efficiency; the ability to constantly shift between goal-directed and habitual strategies, as well as integrate them into behavioral output, is key to optimal performance in everyday life. When such ability is impaired, it may lead to loss of control and to compulsive behavior. Habits have successfully been induced and investigated in rats using methods such as overtraining stimulus-response associations and outcome devaluation, respectively. However, such methods have ineffectively measured habits in humans because (1) human habits usually involve more complex sequences of actions than in rats and (2) of pragmatic impediments posed by the extensive time (weeks or even months), it may take for routine habits to develop. We present here a novel behavioral paradigm—a mobile-phone app methodology—for inducing and measuring habits in humans during their everyday schedule and environment. It assumes that practice is key to achieve automaticity and proficiency and that the use of a hierarchical sequence of actions is the best strategy for capturing the cognitive mechanisms involved in habit formation (including “chunking”) and consolidation. The task is a gamified self-instructed and self-paced app on a mobile phone that enables subjects to learn and practice two sequences of finger movements, composed of chords and single presses. It involves a step-wise learning procedure in which subjects begin responding to a visual and auditory cued sequence by generating responses on the screen using four fingers. Such cues progressively disappear throughout 1 month of training, enabling the subject ultimately to master the motor skill involved. We present preliminary data for the acquisition of motor sequence learning in 29 healthy individuals, each trained over a month period. We demonstrate an asymptotic improvement in performance, as well as its automatic nature. We also report how people integrate the task into their daily routine, the development of motor precision throughout training, and the effect of intermittent reinforcement and reward extinction in habit preservation. The findings help to validate this “real world” app for measuring human habits.
The concept of habit learning has been extensively studied across distinct fields of research, using different methodologies (for a comprehensive review on habits, see
Previously, self-reported questionnaires have been used to investigate aspects of habits distinguishing between routine and automatic tendencies in humans (
“Ecological” paradigms have also been used to track “real-world” habits (
Capitalizing on novel technology, we developed a smartphone motor sequence application to measure habit formation within a more naturalistic setting (at home). Habit strength is promoted here by the permanent accessibility of the app (given that most people carry their mobile phones everywhere), which facilitates training frequency and enables context stability since the tactile, visual, and auditory stimuli associated with the phone and its operation establishes a strong context for all participants regardless of their concurrent circumstances. Thus phone-based tasks favor habit formation since as the frequency of the behavior increases in a stable context, so it increases the strength of the context-behavior association, an effect that is crucial for habit development (
We continuously collected data online, in real time, thus enabling measures of progressive learning and of processes involved in habit formation such as “caching” (
The application incorporated attractive sensory features in a game-like setting, in which participants earned reward points according to their performance (see video for illustration in the “Methods” section). This app-based method for measuring habits in the real world is based on previous findings that have defined training frequency, context stability, and reward contingencies as important for increasing habit strength (
In order to assess the autonomy of habits from goal-directed actions behavioral neuroscientists employ goal devaluation or contingency degradation strategies as interventions to probe habitual control (
Twenty-nine volunteers, recruited from the community
The task consisted of a motor practice program that participants committed to pursue daily, for a period of 1 month (see description of the task design in
Habit Training Task.
Using a simple and self-instructed application downloaded to their mobile devices, participants learned and practiced two sequences of fingers movements, composed of chords (two or three simultaneous finger presses) and single presses (one finger only). Each sequence comprised six moves, performed using four fingers of the dominant hand (index, middle, ring, and little finger). Sequence generation was randomized so that each participant had their own pair of sequences to practice throughout the month. This randomization was conducted to rule out finger-specific effects at individual sequential positions, as each finger will contribute equally to the RTs at each sequential position. However, for each sequence, the order of finger movements was pseudo-randomly generated such that (1) all sequences had three single press moves, two two-finger chord moves, and one three-finger chord move and (2) difficult finger combinations were avoided, for example, a three-finger chord with simultaneous index, middle, and little fingers or index, ring, and little fingers. Therefore, despite being different, all sequences had a similar level of difficulty.
Participants were instructed to respond swiftly and accurately. They were required to keep their fingers very close to the keys to minimize amplitude variation and to enable them to play quickly. To enable sequence learning and memorization, three levels of increased difficulty guided practice. Initially, subjects responded to a visually and auditory cued sequence: they simply followed lighted keys, also associated with musical notes (level 1). These exteroceptive cues were slowly removed throughout the practice progression such that level 2 only included auditory cues and level 3 contained no cues. Successful performance at each stage resulted in progression to the next. Unsuccessful performance resulted in titration to the immediately preceding stage.
Participants received continuous feedback on their performance. Successful trials were followed by a positive ring tone and mistakes by a negative ring tone. Every time a mistake occurred (irrespective of which move in the sequence they were), participants had to restart the sequence in order to perform it entirely correctly.
As previously mentioned, all participants had to practice two motor sequences, each identified by a specific abstract picture. Each sequence was associated with a specific reward schedule. In our design, one of the reward schedules was continuous reward (points were received for every successful trial, as a function of the speed of performance) and the other a variable reward schedule (points were randomly received on 37% of the trials). Calculation of the points was as follows: points decreased linearly from 100 to 0 over 1 second; the counting started as soon as the app became ready to receive the user’s input. This counter reset and restarted counting after each move. As soon as the keys were pressed (for each move), the counter stopped and registered the points achieved for that move. The points received after each sequence was completed were the sum of all the points achieved on each move. All this within-move counting was done in the background so participants only saw the points gained for each sequence once they completed it. This system was implemented to promote speed: the faster participants played the sequence, the more points they gained. If they were too slow, that is, if the key press occurred after the counter had reached 0, then no points were gathered for that particular move. In the continuous reward sequence, participants received the total points acquired after each successful trial completed. In the variable reward sequence, there was a 63% chance that any points earned on a sequence would be set to zero. To compensate for the missing points, the earned points provided on this schedule were doubled. Therefore, both sequences resulted in similar scores by the end of practice. By this point, after 20 sequences had been completed (see “Practice Schedule” section below), subjects could see the total (cumulative) points achieved throughout the practice. While playing, they could also see their current total, gathered at a particular moment. To promote motivation, feedback was also given across daily practice sessions, so subjects could compare their performance across practices and see whether they were improving over days.
All participants were presented with a calendar schedule and were asked to practice both sequences daily (
Thirty days of practice were required, and all data were anonymously collected in real time, through an online server. At the 21st day of practice, the reward schedules were removed (extinction) to test how autonomous of external feedback the response sequence had become. This procedure (1) ensured that the response sequence was more dependent on interoceptive (proprioceptive and kinesthetic) feedback and on the subjects’ internal motivation to continue the training and (2) ensured that we were able to measure and train response sequences triggered by their context, which persisted without explicit reinforcement.
An orientation session, lasting between 30 and 60 min depending on people’s dexterity, was conducted at the Herschel Smith Building, Addenbrooke’s Hospital, in Cambridge. During this session, the researcher helped the participant to download the app to their devices, reviewed the training instructions, and discussed how the task works. All participants were instructed to practice every day to make sure they could perform both sequences automatically and rapidly as they would be assessed in a second session taking place 1 month later. This cover story was introduced in preparation for a follow-up session including a devaluation strategy, which assessed participants’ preferences for habitual sequences over goal-seeking sequences. This task manipulation would test the hypothesis that the behavioral mechanism underlying the transition from a goal-directed to a habitual action is that the action, with repetition, acquires the rewarding properties of its outcome, which may simply be its own proprioceptive/kinesthetic feedback (data to be reported elsewhere).
Behavioral output measures included sequence accuracy and sequence completion times (learning rates), temporal pattern of daily practice, days until habit acquisition, performance as function of different reward schedules, effect of reward extinction and finger position and timings.
For more detailed analyses, we broke down the sequence completion times into two components: (1)
App data were automatically uploaded to a Cloud-based database. Data analysis was performed using custom scripts in MATLAB and Python.
As shown in
App engagement.
We analyzed the two blocks of practice pre- and post-removal of the external rewarding feedback occurring after 21 days of training. After
Significant improvements in accuracy (
Performance.
When decomposing the sequence completion time on successful trials into preparation (i.e., quantifying the time just before a move) and motor-related components (
Decomposing sequence completion times into preparation and motor related components.
We also assessed how motor precision, as measured by finger position variance, varied throughout training. This measure was computed using the X and Y pixel coordinates of participants’ screen touches (
Motor precision.
We have presented an experimental paradigm based on motor sequence learning which can be employed to study, in a systematic and controlled way, the building blocks of more complex behavioral sequences that make up our everyday real-world actions. Designed as a smartphone tool, and thus easily available to subjects, it enabled for the first time, the induction and measurement of habitual behavior in humans during their everyday schedule, routines, and environment (in the comfort of their homes), while collecting continuously 30 days of real-time data. Such a naturalistic experimental set-up may perhaps be useful for the future investigation of habits.
The test paradigm is assumed to encompass multiple and continuous cycles of model-free and model-based learning processes thought to be required for habit development, which include processes of instrumental or operant reinforcement, adaptation, plasticity, and other explicit cognitive processes (
This app-based method to measure habits in the real world is based on previous literature which has isolated frequency, context stability, rewards, and simplicity as important factors that promote habit strength (
Automaticity was measured in terms of three criteria: sequence completion times, progressive extinction of learning cues, and autonomy from the goal as assessed by extinction. Additionally, proficiency was also measured in terms of motor precision. The significant increase in finger variance throughout the training is also a strong indicator of motor performance optimization. According to optimal feedback control theory (
In terms of proficiency and automaticity, sequence completion times significantly improved throughout training, reaching asymptotic performance levels between practice blocks 40 and 50. The exponential decay in error rates to an asymptote and further optimization of the speed/accuracy trade-off is clear evidence of learning and skill development. The greater improvement in sequence completion time during the initial 20 blocks corresponds to the “fast learning” mode, typically observed during the goal-based acquisition phase mediated by the associative striatal regions, in coordination with cerebellum, prefrontal, and premotor cortical regions (
The preservation of this skilled behavior after extinction of the external cues, maintaining the same high level speed-accuracy trade-off, is an additional sign of automaticity and habitual control. Our findings are consistent with
One test of habitual control effected in this task was extinction, involving the omission of explicit reward feedback. The removal of rewarding feedback on the 21st day of training mainly affected errors. Although there was a small effect on sequence completion time (only in the practice condition), this was much less significant, possibly indicating that performance had indeed attained a degree of autonomy from the goal. This suggests that the motor sequence had become habitual in part but still retained some sensitivity to goal despite extinction (and hence goal-directed control). Of course, this extinction manipulation did not remove all forms of motivation from performance because of the degree of intrinsic motivation that humans exhibit in such research studies.
Although one could expect different learning patterns as consequence of different reward schedules, we did not observe significant effects of the reward feedback schedule (i.e., continuous versus variable) on habit development. There was, however, a selective effect of reward schedule in performance during the switch test. The detrimental effect of extinction on this switch test depended on the previous schedule of reward feedback, specifically occurring in the continuous condition only. A possible explanation for this might be that pitting two habits against one another in an explicit choice situation recruits executive processing, hence re-engaging the goal-directed system, which may be more vulnerable to extinction in the continuously rewarded condition because the change in reward contingency is more immediate and explicit than for the variable schedule. Future studies may seek to vary the nature of intermittency of the reward schedule by explicitly comparing random ratio versus random interval schedules, the latter being associated with greater habitual control (
This study has a few limitations and challenges to consider. Its ecological nature, enabling people to conduct the task in the comfort of their homes, including it in their everyday schedule, routines, and environment and at their own pace, partly solves the major problem of the artificial nature of previous studies. However, this feature limits the study to its behavioral nature, making it more difficult if one wants to investigate the neural basis of habit and skill development using functional imaging. There were also some technical difficulties to deliver the app on android phones, confining our recruitment to Apple users, which obviously decreased our recruitment pool of subjects. Several iPods were purchased for lending to participants in order to facilitate recruitment. Additionally, the study required careful monitoring by the researchers on a daily basis, to track participants’ commitment, gauge motivation, and send reminders when needed. Conducting this study with clinical populations might be challenging, given that some patients may not be so motivated as healthy volunteers. However, this concern has not in fact been the case with patients with OCD we have also begun to recruit, following the same procedures. Despite all the technical challenges, which also included a continuous update of the online server for data collection, such an advanced methodology was worth pursuing since it facilitated the acquisition of a large dataset, without requiring much effort from our participants.
In conclusion, this article aimed to validate a novel behavioral method for measuring motor response sequence habits in the real world using a mobile phone app. The analysis provided here is preliminary but sufficient to show that proficiency and automaticity is attained according to several different criteria. When tested in clinical populations, the method may provide new insights into the mechanisms underlying abnormal habit learning and corticostriatal functioning in psychiatric disorders and their putative contributions to compulsive behavior. Ongoing research is using this novel app to investigate the neural mechanisms of compulsive behavior in patients with OCD. More generally, this app-based approach could be deployed in a wide variety of discrete sequential production paradigms including dexterity, music, and memory training. It could also be used in the studies of individual differences, for example, to investigate whether aspects of the Big 5 predict how quickly habit/skill is developed.
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.
All participants were given a letter of information, gave written informed consent prior to participation, in accordance with the Declaration of Helsinki, and were financially compensated for their participation. This study was approved by the Cambridge South Research Ethics Committee.
PB and TR conceived the idea and designed the method. PB and TP developed the app. PB carried out the experiment. PB and DM conducted the data analyses. QL helped with the data preparation. PB, DM, and TR wrote the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We would also like to thank the intern students Samantha De La-Rocque, Emilia Szmyrgala, and Rebecca Richards for their help piloting the task. PB is a postdoctoral research fellow at Hughes Hall College, Cambridge and during the revision of this manuscript QL was a Visiting Fellow at Clare Hall, Cambridge, UK.