Moving beyond response times with accessible measures of manual dynamics

Button-press measures of response time (RT) and accuracy have long served a central role in psychological research. However, RT and accuracy provide limited insight into how cognitive processes unfold over time. To address this limitation, researchers have used hand-tracking techniques to investigate how cognitive processes unfold over the course of a response, are modulated by recent experience, and function across the lifespan. Despite the efficacy of these techniques for investigating a wide range of psychological phenomena, widespread adoption of hand-tracking techniques within the field is hindered by a range of factors, including equipment costs and the use of specialized software. Here, we demonstrate that the behavioral dynamics previously observed with specialized motion-tracking equipment in an Eriksen flanker task can be captured with an affordable, portable, and easy-to-assemble response box. Six-to-eight-year-olds and adults (N = 90) completed a computerized version of the flanker task by pressing and holding a central button until a stimulus array appeared. Participants then responded by releasing the central button and reaching to press one of two response buttons. This method allowed RT to be separated into initiation time (when the central button was released) and movement time (time elapsed between initiation and completion of the response). Consistent with previous research using motion-tracking techniques, initiation times and movement times revealed distinct patterns of effects across trials and between age groups, indicating that the method used in the current study presents a simple solution for researchers from across the psychological and brain sciences looking to move beyond RTs.

& Eriksen, 1974). For instance, Erb and Marcovitch (2018) found that movement initiation times (the time elapsed from stimulus presentation to movement initiation) revealed a different pattern of trial sequence effects than movement times (the time elapsed from movement initiation to response completion) and reach curvatures (a measure of the extent to which a movement trajectory deviated from a direct path to the selected response target). Further, the researchers observed different developmental trajectories in these measures, indicating that key processes underlying cognitive control develop at different rates. The current study will address three main questions. Q1. Can the manual dynamics observed with motion-tracking equipment be captured using more portable and cost-effective equipment? The electromagnetic position and orientation recording system used by Erb and colleagues (Erb & Marcovitch, 2018;Erb, Moher, Sobel, & Song, 2016) is not portable and is relatively expensive. The equipment is also sensitive to ferromagnetic metals, making certain types of data collection difficult (e.g., gathering eye-tracking and hand-tracking data simultaneously). Thus, a primary aim of this project is to identify whether a button-release-and-press method can be used to capture the manual dynamics observed with the electromagnetic position and orientation recording system. To that end, we developed a response box featuring a "home" button at the bottom centre of the box and two "response" buttons located toward the top corners of the box.
To begin each trial, participants will press the "home" button. After the imperative stimulus array is presented, participants will release the "home" button and press one of the two "response" buttons. This will enable us to measure movement initiation time (the time elapsed between stimulus presentation and when the "home" button is released), movement time (the time elapsed between when the "home" button is released and when a "response" button is pressed), and response time (the time elapsed between stimulus presentation and when a "response" button is pressed). We will also use a video camera mounted over the response box to record hand movements. Our goal is to use a software package (e.g., Kinovia or OpenPose) to extract movement trajectories for each trial and then use these trajectories to compute curvature values in the same manner as Erb and colleagues (Erb & Marcovitch, 2018;Erb et al., 2016). To the extent that the manual dynamics observed with this button-release-and-press technique match those observed with less portable and less cost-effective motion-tracking techniques, the button-release-and-press approach might serve as an effective technique for researchers looking to incorporate measures of manual dynamics into their research. For example, the technique could be especially helpful for researchers conducting research outside of laboratory settings (e.g., in museums or schools Erb and Marcovitch (2018) revealed age-related gains in flanker task performance between middle childhood and adulthood in initiation times, movement times, response times, curvatures, and error rates. This study will evaluate the extent to which these developmental differences are observed with a button-release-and-press method. The study will also explore the extent to which the age-related gains in flanker performance might reflect a transition occurring during middle childhood from a more reactive style of control to a more proactive style of control (Chevalier, 2015, Child Development Perspectives). For example, in the flanker task, children who adopt a more reactive style of control may not adequately prepare for the upcoming trial by maintaining a constrained focus on the cue that appears before the stimulus array or the target arrow that appears in the stimulus array surrounded by distractor arrows.
One way to assess this possibility is by measuring looking behaviour and assessing the extent to which looking behaviour in the flanker task differs across development. Similarly, at an individual differences level, one could also evaluate whether individuals who do maintain a constrained focus on the cue and target outperform those who do not.

Hypotheses
Hypotheses regarding Q1: If the button-release-and-press technique is capable of capturing the same manual dynamics observed in reaching behaviour, then performance on the flanker task should replicate the patterns of trial sequence effects observed in initiation time, movement time, response time, and curvature by Erb and Marcovitch (2018).
Hypotheses regarding Q2 and Q3: If the button-release-and-press technique is capable of capturing the same developmental dynamics observed in reaching behaviour, then performance on the flanker task should reveal similar age-related differences to those observed between children and adults by Erb and Marcovitch (2018). If the age-related gains in flanker performance observed across childhood and between childhood and adulthood reflect a transition from a more reactive mode of control to a more proactive form of control, we should observe age-related differences in eye fixations during the presentation of the cue and during the presentation of the target and distractors. Specifically, children should fixate on the cue and target less than adults and should fixate on task irrelevant features (e.g., distractors or other parts of the screen) more.

Study type
Experiment -A researcher randomly assigns treatments to study subjects, this includes field or lab experiments. This is also known as an intervention experiment and includes randomized controlled trials.

Blinding
No blinding is involved in this study.

Study design
The study will be a mixed design. The within-subject factors will be the congruency of the current trial (2; congruent vs. incongruent), the congruency of the previous trial (2; congruent vs. incongruent), and the response-repetition type of the current trial (2; Does the response generated on the current trial repeat or alternate relative to the previous trial?). The betweensubjects factor will be age group (2; children vs. adults).

Randomization
Participants will not be assigned randomly to specific conditions. The presentation of trials will be randomized for each participant, meaning that one of four stimulus arrays (congruent with left response, congruent with right response, incongruent with left response, incongruent with right response) will be randomly presented on each trial, with each block consisting of an equal number of each stimulus array.

Existing Data
We are currently in the process of collecting data for the study but we have not yet processed or analysed any of the collected data. At of the time of this pre-registration, we have collected 11 child participants and 8 adult participants. Task: Participants will perform a version of the Eriksen flanker task based off of the task used by Erb and Marcovitch (2018, Cognition). Participants will first undergo a calibration phase with the eye-tracking system. Next, an experimenter will describe the rules of the flanker task and introduce participants to the button-release-and-press method. Participants will complete a set of practice trials. Following the practice trials, participants will be presented with 4 blocks of 48 trials. These blocks will consist of 24 congruent trials and 24 incongruent trials.

Data collection procedures
Half of the trials will require a left response, whereas the other half will require a right response. The presentation of trials will be randomised in each block. During performance of the task, hand movements will be video recorded from a camera mounted over the response box. After the flanker task is complete, child participants will perform a standardised version of the Heads, Toes, Knees, and Shoulders (HTKS) task (Ponitz, McClelland, Jewkes, Connor, Farris, & Morrison, 2008).

Sample size
Our target sample size is 90, with 45 child participants and 45 adult participants.

Sample size rationale
Our sample size for each age group was selected to match that of Erb and Marcovitch (2018, Cognition).

Stopping rule
We will stop data collection once we have collected usable data from 45 child participants and 45 adult participants (see Data Exclusion and Missing Data sections for details regarding what data will count as usable). If, during the course of processing and analysing the data, we discover an issue with the task, program, or equipment that substantially impairs our ability to address the three research questions of interest, we will stop data collection.

Manipulated variables
We will have three within-subjects factors: (1) the congruency of the current trial (C vs. I), (2) the congruency of the previous trial (c vs. i), and (3) the response repetition type of the current trial (response repetition vs. response alternation). We will have one between-subjects factor: (1) age group (children vs. adults).

Measured variables
Manual Measures: (1) Response time (time elapsed from stimulus presentation to response completion) (2) Initiation time (time elapsed from stimulus presentation to the release of the "home" button) (3) Movement time (time elapsed from the release of the "home" button until response completion) (4) Curvature (a ratio score reflecting the degree to which a movement trajectory deviated from a direct path to the selected target). Note: we will be attempting to use software packages to extract the spatial characteristics of hand movements from video files. It is presently unclear how effective this approach to computing reach curvatures will be.
It is also unclear how much time will be needed to extract the relevant parameters. It is therefore possible that this measure will not be computed and analysed. Note: Our attempts to extract the spatial characteristics of hand movements from the video files was unsuccessful, due to difficulties with automatising the process of identifying the location of the hand in each frame.
Oculomotor measures: (1) Cue fixation proportion score: The number of samples with fixations on the cue divided by the total number of samples gathered during the cue presentation period.
In other words, for each trial, we will be identifying how many of the collected fixation measurements were on the cue. If fixation data were unavailable for samples (e.g., due to dropped samples or the participant looking off screen), these samples will not contribute to the proportion score.
(2) Target fixation proportion score: The number of samples with fixations on the target divided by the total number of samples gathered during the presentation of the target and distractors. In other words, for each trial, we will be identifying how many of the collected fixation measurements were on the target during the presentation of the target and distractor items. If fixation data were unavailable for samples (e.g., due to dropped samples or the participant looking off screen), these samples will not contribute to the proportion score.
(3) Distractor fixation proportion score: The number of samples with fixations on any of the distractors divided by the total number of samples gathered during the presentation of the target and distractors. In other words, for each trial, we will be identifying how many of the collected fixation measurements were on one of the distractors during the presentation of the target and distractor items. If fixation data were unavailable for samples (e.g., due to dropped samples or the participant looking off screen), these samples will not contribute to the proportion score.

Indices
We will be computing mean performance for each trial type and for each participant in the same manner as Erb and Marcovitch (2018, Cognition).

Statistical models
With regard to our measures of hand movements and eye fixations, our analytic strategy will mirror that of Erb and Marcovitch (2018, Cognition). We will use ANOVAs to evaluate the effects of Age Group, Current Congruency, Previous Congruency, and Response Repetition Type on the measures outlined above. We will follow-up on interaction effects in the same manner as Erb and Marcovitch (2018, Cognition) (note: the primary difference will be that Erb and Marcovitch collected data from three age groups, whereas the current study will collect data from two age groups). In addition to these analyses, we will evaluate the correlation between our oculomotor measures and age in months for our child participants to evaluate whether eye movements show evidence of a shift from a more reactive mode of control to a more proactive mode of control. We will also evaluate the correlation between our oculomotor measures and the size of the congruency effect observed in response times for our child participants. Finally, we will evaluate the correlation between the size of the congruency effect and performance on the HTKS task for child participants in order to identify the extent to which the measures might be tapping similar inhibitory control abilities.

Transformations
Following Erb and Marcovitch (2018, Cognition), we will log-transform the initiation time, movement time, response time, and reach curvature data. We will also run our analyses with the non-transformed data and make the results of those analyses available either in the main text or supplementary materials of any resulting publication. We will not log-transform the eye movement data.

Inference criteria
Following Erb and Marcovitch (2018, Cognition), we will set our alpha value to .05 for all analyses. All analyses will be two-tailed. Bonferroni corrections will be applied to adjust for multiple comparisons in both sets of correlational analyses.

Data exclusion
To be included in the final sample, participants will need to have completed (and have usable data from) at least 3 out of the 4 blocks of experimental trials. For all analyses, the first trial of each block will be excluded. For all analyses except for our analysis of error rates, only trials featuring an accurate response and following an accurate response will be included.

Missing data
We anticipate that eye movement data will not be available for each trial, as participants (particularly child participants) often move their head out of the region monitored by the eye tracker or obscure their eyes (e.g., by resting their head in their hand). If a participant's eye movement data are not available for at least 50% of the completed experimental trials, the eye tracking data for that participant will not be included in our analyses. Note that the participant's behavioural data will still be analysed, even if their eye movement data are not.
In other words, the participant will still count towards the target sample size of 45 participants in each age group. In the event that the video of a participant's reaching behaviour is unavailable (e.g., due to technical difficulties), the participant's behavioural data (not including curvature) will still be analysed as well.

Exploratory analysis
We may perform exploratory analyses regarding pupillometry measures. As part of this project, we are also testing a smaller sample of 5-year-olds. In pilot testing, we found that 5- year-olds had a difficult time completing three or more blocks of the experimental trials.
However, we are interested in evaluating 5-year-olds' performance on the task, particularly with regard to their looking behaviour. We will therefore analyse 5-year-olds' data (assuming that at least two blocks of experimental trials have been completed) in the same manner as older children's and adults'. Finally, we will explore performance on trials featuring an error to identify, for example, whether fixation patterns differed on error trials relative to non-error trials. For instance, were fixations on the cue lower on error trials relative to non-error trials?
Were fixations on the distractors higher?
Note: In the subsequent supplementary materials, we will not include data collected from the 5-year-old sample, in-keeping with the results reported in the current manuscript.

Section 2: Pre-registered analyses
In-keeping with our pre-registration, the following analyses pertaining to eye movement include data from participants with available eye movement data for over 50% of the experimental trials. Five children (two 6-year-olds, one 7-year-old, and two 8-year-olds) did not meet this criterion and were excluded accordingly.

HTKS scores
We did not find a significant correlation between HTKS score and the size of the congruency effect observed in response times for our child participants r(39) = .11, p = .491.

Background
The development of open-source electronics with easy-to-use hardware and software has allowed many researchers, including those with little to no prior knowledge of electronics and programming, to construct simple instruments for use in research.
For example, USB-based microcontrollers (e.g., Teensy, Arduino) can be wired with push buttons and programmed as a simple keyboard device for use in recording responses in cognitive tasks. Commonly used software in psychological research, such as E-Prime©, are typically compatible with keyboard input. These devices can therefore be readily incorporated into existing program designs. Components can be acquired at very low cost, and easily assembled to suit a wide range of experimental designs and response configurations. Moreover, with no licensing or software constraints, these devices can be used with open-source packages such as PsychoPy.
The construction of a basic response box can be divided into three steps. The second step involves wiring the electrical components. The level of electrical work required to wire the button box is comparable with that found in consumer-level electrical hobbyist kits. Many universities provide access to resources that allow for basic training on electronics and facilities to carry out this type of work. Where this is not available, online communities provide a valuable source of information and guidance, and the basic tools required for the level of electrical work can be acquired at a minimal cost.
The third step requires the programming of the microcontroller to carry out its required functions. Guidance on this can be found on the manufacturer's website as well as various online communities and platforms. Programming a microcontroller as a keyboard device is entry-level and requires very little prior knowledge or experience. Existing code can be used or modified as required.
In the following, we describe a simple method for constructing and programming a buttonbox capable of capturing response dynamics typically observed with more complex optical or electromagnetic motion capture equipment. The button-box comprises three simple push buttons wired to microcontrollers, which record the press and release of each button. The button-box can be connected to a computer using the built-in USB cable, and is compatible with any software capable of receiving input from a keyboard. Our design can be easily adapted to include more response options or configurations, making it an accessible and adaptable solution for researchers looking to design their own purpose-built response apparatuses.

A guide to construction
This response box was developed using a Teensy microcontroller and behaves as a 3-key keyboard. A Takashi CF-series sloped enclosure was fitted with three white Sanwa pushbuttons, and programmed using open-source software, Arduino IDE 1.8.9. A built-in library allows for keyboard functions to be readily programmed into the Teensy board, enabling the buttons to send "keystrokes" to a computer via the microcontroller board's native USB port. An open-source third party library, Bounce2 (https://github.com/thomasfredericks/Bounce2) was also used. Note: Bouncing occurs when any two metal contacts in an electronic device, such as a button, make contact and generate a signal. Due to their mechanical nature, a single button press or release can yield multiple signals. To control for this, debouncing software is often used to ensure one button or key press/release results in one signal being sent and registered by the computer. Step 1: Enclosure 1. Mark the locations for each button on the top panel of the enclosure 2. Drill a 30mm hole in each location and insert the buttons so that the connects of the buttons align left and right 3. Drill a hole large enough for the USB cable to pass through on either a side panel or back panel of the enclosure 4. Build enclosure as per manufacturer instructions, leaving off the top panel and one side panel Step 2 Wire from ground to the right leg of the right button 4. To complete the grounding connections, solder: a. A wire connecting the right leg of the right button to the right leg of the centre button b. A wire connecting the right leg of the centre button to the right leg of the left button. 5. Thread the USB cable through the hole and plug it into the teensy microcontroller. 6. Slide the top panel into place, and attach the side panel a. Note: you may want to secure the Teensy board to the top of the panel with hot glue, so that wires don't become loose when the box is being transported. It may be best to do this once the board is programmed and you are certain everything is working as it should.
Step 3 A small window with a picture of the teensy board will appear, select Operation > Program. When the code is uploaded, it will show a message saying "download complete". 13. Open a notepad and test the buttons, making sure the letters are matched to the correct keystrokes ("r" for the right button, "l" for the left button, SPACE for centre)   The following code can be used to program a simple 3-button response box using a Teensy++ 2.0 micro controller. Internal pull-up resistors are used for each button.
Bounce 2 library is used to filter mechanical noise from the buttons. The Debouncer function introduces a timing threshold that a signal must meet for the Teensy++ 2.0 to register it as a button/key press or release. We used a threshold of 5ms, such that the button must be pressed for 5ms for the signal to be processed as a button press/release and sent to the PC. Section 5: Analyses of raw data