Motor Sequencing App.

a) A trial starts with a static image depicting the abstract picture that identifies the sequence to be performed (or ’played’) as well as the 4 keys that will need to be tapped. Participants use their dominant hand to play the required keys: excluding the thumb, the leftmost finger corresponds to the first circle and the rightmost finger corresponds to the last circle. b) Screenshot examples of the task design: (1) sequence selection panel, each sequence is identified by an abstract picture; (2) panel exemplifying visual cues that initially guide the sequence learning; (3) panel exemplifying the removal of the visual cues, when sequence learning is only guided by auditory cues. c) Example of a sequence performed with the right hand: 6-moves in length, each move can comprise multiple finger presses (2 or 3 simultaneous) or a single finger press. Each sequence comprises 3 single press moves, 2 two-finger moves, and 1 three-finger move. d) Short description of the daily practice schedule. Each day, participants are required to play a minimum of 2 practices per sequence. Each practice comprised 20 successful trials. Participants could play more if they wished and the order of the training practices was self-determined.

Training Engagement.

a) Whole training overview. OCD patients engaged in significantly more training sessions than HV (* p = 0.005). The minimum required practices (P) were 120P. b) Daily training pattern for HV and c) Daily training pattern for OCD. Single dots on the unit circle denote the preferred practice times of individual participants within 0-24 hours, obtained from the mean resultant vector of individual practice hours data (Rayleigh statistics). Group-level statistics was conducted in each group separately using the Rayleigh test to assess the uniformity of a circular distribution of points. The graphic displays the length of the mean resultant vector in each distribution, and the associated p-value. Regarding between-group statistical analysis, see main text.

Learning.

Upper panel: Model fitting procedure conducted for the continuous reward sequence. Lower panel: Model fitting procedure conducted for the variable reward sequence. a) Individual plots exemplifying the time-course MT (in seconds) as training progresses (lighter color) as well as the exponential decay fit modeling the learning profile of a single participant (darker color). Left panels depict data in a HV individual, right panels display data in a patient with OCD. b) Group comparison resulting from all individual exponential decays modeling the learning profile of each participant. A significant group difference was observed on the amount of learning, MTL, in both reward schedule conditions (continuous: p = 0.009; variable: p < 0.001). Solid lines: median (M); Transparent regions: median +/- 1.57 * interquartile range/sqrt(n); Purple: healthy volunteers (HV); Blue: patients with obsessive-compulsive disorder (OCD).

Automaticity.

a) We mathematically defined trial-to-trial inter-keystroke-interval consistency (IKI consistency), denoted as C (in seconds), as the sum of the absolute values of the time lapses between finger presses across consecutive sequences. The variable n represents the sequence trial and k denotes the IKI. We evaluated automaticity by analyzing the decline in C over time, as it approached asymptotic levels. b) Group comparison resulting from all individual exponential decays modeling the automaticity profile (drop in C) of each participant. A significant group effect was found on the amount of automaticity gain, CL(Kruskal–Wallis H = 11.1, p < 0.001) and on the automaticity constant, n) (Kruskal– Wallis H = 4.61, p < 0.03). Solid and dashed lines are median values (M). Light purple: healthy volunteers (HV); Dark purple: patients with obsessive-compulsive-disorder (OCD); Solid lines: continuous reward condition; Dashed lines: variable reward condition.

Sensitivity of movement time to changes in reward in the continuous reward schedule.

a) Mean normalized change in movement time (MT, ms) from trial n to n+1 following an increment (ΔR+, in purple) or decrement (ΔR-, in green) in scores at n. The change in movement time trial-to-trial was normalized with the baseline value on the initial trial n: ΔMT(n+1) = (MT(n+1) - MT(n))/MT(n). This relative change index is therefore adimensional. The dots represent mean MT changes (error bars denote SEM) in each bin of correctly performed sequences, after partitioning all correct sequences into four subsets, and separately for OCD (dark colors), and HV (light colors). b) Both groups of participants speeded up their sequence performance more following a drop in scores (main effect of reward, p = 2 x 10-16 ; 2 x 4: reward x bin ANOVA); yet this acceleration was reduced over the course of practiced sequences (main bin effect, p = 5.06 x 10-12). c) Same as a) but for the spread (std) of the MT change distribution (adimensional). d-e) Illustration of the main effect of group ([d] p = 9.93 x 10-6) and reward ([e] p = 4.13 x 10-5) on std. Each bin depicted in the plots (x-axis) contains around 110 correct sequences on average (further details in Supplementary Results: Sample size for the reward sensitivity analysis).

Sensitivity of normalized IKI consistency (normC) to reward changes in the continuous schedule.

a) The mean normalized change in trial-to-trial IKI consistency (normC, equation [8]; adimensional) across bins of correct sequences is shown, separately for each group (OCD: dark colors; HV: light colors) and for reward increments (ΔR+, purple) and decrements (ΔR-, green). The dots represent the mean value, while the vertical bars denote SEM. b) Illustration of the main effect of group (left panel; p = 0.00454) and type of reward change (right panel; p = 1.86 x 10-06). c) Same as a) but for the std of the distribution of IKI consistency changes, normC, adimensional. d) The panel displays the main effect of bin (p = 3.63 x 10-14) on the std. Black denotes the average (SEM) across reward and group levels. Each bin depicted in the plots (x-axis) contains 110 correct sequences on average (See Supplementary Results: Sample size for the reward sensitivity analysis).

Preference for familiar versus novel action sequences.

a) Explicit Preference Task. Participants had to choose and play one of two given sequences. Once choice was made, the image correspondent to the selected sequence was highlighted in blue. Participants then played the sequence. While playing it, the bar on top registered each move progressively lighting up in green. There were 3 conditions, each comprising a specific sequence pair: 1) app preferred sequence versus app non-preferred sequence (control condition) 2) app preferred sequence versus any 6-move sequence of participant’s choice (experimental condition 1); 3) app preferred sequence versus any 3-move sequence of participant’s choice (experimental condition 2). b) No evidence for enhanced preference for the app sequence in either HV nor OCD patients. In fact, when an easier and shorter sequence is pitted against the app familiar sequence (right raincloud plot), both groups significantly preferred it (Kruskal-Wallis main effect of Condition H = 23.2, p < 0.001). Left raincloud plot: control condition; Middle raincloud plot: experimental condition 1; Right raincloud plot: experimental condition 2. Y-axis depicts the number of app-sequence choices (15 choice trials maximum). Connected lines depict mean values. (c) Exploratory analysis of the preference task following up unexpected findings on the mobile-app effect on symptomatology: re-analysis of the data conducting a Dunn’s post hoc test splitting the OCD group into 2 subgroups based on their YBOCS change after the app training [14 patients with improved symptomatology (reduced YBOCS scores) and 18 patients who remained stable or felt worse (i.e. respectively, unchanged or increased YBOCS scores)]. Patients with reduced YBOCS scores after the app training had significantly higher preference to play the app sequence in both experimental conditions (left panel: pFDR = 0.015*; right panel: pFDR = 0.011*). The bar plots represent the sample mean and the vertical lines the confidence interval. Individual data points are included to show dispersion in the sample. Abbreviations: YBOCS = Yale-Brown obsessive-compulsive scale, HV = Healthy volunteers, OCD = patients with obsessive-compulsive disorder.

Re-evaluation procedure: 2-choice appetitive learning task.

a) shows the task design. We tested 4 conditions, with chest-pairs corresponding to the following motor sequences: 1) app preferred sequence versus any 6-move sequence; 2) app preferred sequence versus novel (difficult) sequence; 3) app preferred sequence versus app non-preferred sequence; 4) app preferred sequence versus any 3-move sequence. The ‘any 6-move’ or ‘any 3-move’ sequences could comprise any key press of the participant’s choice and could be played by different key press combinations on each trial. The ‘novel sequence’ (in 2) was a 6-move sequence of similar complexity and difficulty as the app sequences, but only learned on the test day (therefore, not overtrained). In conditions 1, 2 and 3, the preferred app sequence was pitted against alternative sequences of higher monetary value. In condition 4, the intrinsic value of the preferred app sequence was pitted against a motor-wise less effortful sequence (i.e. a shorter/easier sequence). Each condition addressed specific research questions, which are detailed in the right column of the table. b) demonstrates the task performance per group and over the 4 conditions. Both groups were able to adjust to the new contingencies and choose the sequences the sequences associated to higher monetary reward. When re-evaluation involved an motor effort manipulation, OCD patients chose the app sequence significantly more than HV (* = p < 0.05) (condition 4). Y-axis depicts the number of app-sequence chests chosen (40 trials maximum) and connected lines depict mean values.

Mobile-app effect on symptomatology.

a) Left: positive correlation between patients’ routine tendencies reported in the Creature of Habits (COHS) questionnaire and the symptom improvement (Pearson r = 0.45, p = 0.01). Symptom improvement was measured by the difference in YBOCS scale before and after app training. Right: Patients with greater improvement in their symptoms after the one month app training had greater habitual tendencies as compared to HV (p < 0.001) and to patients who did not improve post-app training (p = 0.002). The bar plot represents the sample means and the vertical lines the confidence interval. Individual data points are included to show dispersion in the sample. b) OCD patients who related their symptom improvement directly to the app training were the ones with higher compulsivity scores on the OCI (Pearson r = 0.8, p = 0.008) (left) and higher habitual tendencies on the COHS (Pearson r = 0.77, p < 0.01) (right). Note that b) has one missing patient because he did not complete the OCI and COHS scales. Abbreviations: OCI = Obsessive-Compulsive Inventory, COHS = Creature of Habits Scale, YBOCS = Yale-Brown obsessive-compulsive scale, HV = Healthy volunteers, OCD = patients with obsessive-compulsive disorder.

a) Participants’ demographics and clinical characteristics. b) Between group results from the self-reported questionnaires. Abbreviations: HV, Healthy Volunteers; OCD, Patients with Obsessive-Compulsive Disorder; Y-BOCS, Yale-Brown obsessive-compulsive scale; MADRS, Montgomery-Asberg Depression Rating Scale; STAI, The State-Trait Anxiety Inventory; BDI, Beck Depression Inventory; OCI, Obsessive-Compulsive Inventory; CPAS, Compulsive Personality Assessment Scale; COHS, Creature of Habit Scale; HSCQ, Habitual Self Control Questionnaire; BIS, Behavioral Inhibition System; BAS, Behavioral Activation System; Barratt, Barratt Impulsiveness Scale; IUS, Intolerance of Uncertainty Scale; SCS, Self-Control Scale; FMPS, Frost Multidimensional Perfectionism Scale; PSS, Perceived Stress Scale; PSWQ, Penn State Worry Questionnaire. ** = p < 0.01, *** = p <0.001.

Self-reported measures on various scales measuring impulsiveness, compulsiveness, habitual tendencies, self-control, behavioral inhibition and activation, intolerance of uncertainty, perfectionism, stress and the trait of worry.

Demographic and clinical characteristics of OCD patients and matched healthy controls

Follow up task instructions