On the Use and Benefits of Electronic Data Collection Systems: A Tutorial on Countee

Abstract

Objectivity in behavior measurement is a defining feature of behavior analysis. With the increased accessibility of technology, there has been an increase in electronic data collection methods, which carry several advantages, such as ease of data transformation and transfer from electronic outputs (e.g., Microsoft Excel files). Countee is an electronic data collection application that has been named in over 30 articles in behavior-analytic journals. It is available for free on both iOS- and Android-based smartphones in multiple countries. The application allows for the programming of both frequency and duration keys, which can be used to derive additional measures. Despite its use in behavior analytic research and practice, a tutorial has not yet been published on the application. Therefore, the purpose of this article is to outline its features, provide instructions on its use, data transformations and calculations, and describe its benefits to ease replication and disseminate its use.

Behavior analysis, as a science, is partially defined by its objectivity in measurement, which often includes direct observation of behavior in both research and practice (Baer et al., 1968). Electronic (e.g., applications) and automatic (e.g., heartrate monitors) data collection systems are becoming increasingly popular to record the various dimensions of behavior (e.g., duration, frequency, latency) during either in-vivo or video observations or proxy measures (e.g., Bak et al., 2021). This may be due to some potential advantages over pen and paper methods (Sleeper et al., 2017; Tapp et al., 2006). Although using a smartphone or computer to take data might be more expensive and require additional staff training, it will likely save time, reduce the risk of errors that can occur when transferring data from paper to an electronic format (e.g., Excel worksheet), and decrease the effort required to share data (Tapp et al., 2006). Collecting data via electronic systems might also make it easier to collect data on either multiple dimensions of behavior, such as both frequency and duration, or the behavior of multiple individuals at the same time (e.g., Bullock et al., 2017). Collecting duration-based data might also be easier via electronic systems because it eliminates the need for additional materials that are typically built into applications (e.g., timers). Data collected on many applications are often directly available in a CSV or Microsoft Excel file for analysis, which eliminates the need to manually transfer the data from one source to another (e.g., Bullock et al., 2017; van der Marel et al., 2022). Finally, because data are already in electronic formats, it often makes data transformation (e.g., reorganizing, visual analysis) or performing other data calculations easier (e.g., calculating percentages).

Fortunately, there are several electronic data collection applications that can be downloaded for immediate use on to computers, phones, and tablets, many of which are free of charge. One of these is BDataPro, which was developed by behavior analytic researchers at the Kennedy Krieger and the Monroe-Meyer Institutes. It is currently free of charge, allows for the collection of both frequency and duration of behavior on personal computers; it is not available for use on phones or tablets (Bullock et al., 2017). BehaviorSnap is available to download on tablets, phones, and computers and can also be used to take data on both frequency and duration of behavior. Tutorials on BehaviorSnap application’s use are available on the website; however, one drawback with this application is that it currently costs $9.99 to download a trial version, and then $4.99/month, $14.99/quarter, $49.99/year or $499 for a lifetime subscription. Van der Marel et al. (2022) also reviewed and provided brief tutorials of six electronic data collection systems specifically designed for behavioral observation of animals, although more than one (e.g., Animal Behaviour Pro) has been used to collect data on human behavior. Four of the six are currently free to download and use, but not all are available on all platforms. There are also other data collection systems (e.g., CentralReach) that are designed, not only for observational data, but to manage curricula across learners. These systems can be costly and are typically designed to be used across clinics and centers. Finally, there is Countee, which is the focus of this article (Peic-Gavran & Hernández Eslava, 2020). Like many others, it is free to download on both iOS- and Android-based smartphones in multiple countries and does not require internet access while it is in use. Although a more extensive comparison between these applications is outside of the scope of this article, those who are interested could review the cited tutorials to determine, which, if any, best fit their needs.

The exact degree to which Countee is being used by behavior analysts is currently unknown; however, its use in applied research has been documented in several articles across behavior-analytic journals with the number of references increasing with time. Searching for “Countee” in four major applied behavior analysis (ABA) journals resulted in 37 articles published between the years 2017 and 2023, including three in Behavior Analysis: Research and Practice, seven in Behavior Analysis in Practice, 12 in Behavioral Interventions, and 15 in the Journal of Applied Behavior Analysis. The range of topics of these articles include, but are not limited to, functional analysis (Slanzi et al., 2022), physical activity (Goldman & DeLeon, 2022), task engagement (Hardesty et al., 2023), pediatric feeding (O’Hara et al., 2023), and staff training (Ruby & DiGennaro Reed, 2022).

The consistent increase in use of Countee as documented in ABA journals indicates that a tutorial on the application might be beneficial to both researchers and practitioners. To our knowledge, there are no publications describing its features, instructions or benefits of its use, or examples of possible applications. The lack of a published detailed description of Countee might prevent it from being more widely adopted in research and practice in behavior analysis and other fields. The purpose of this article is to fill this gap in the literature by (1) describing the features of the Countee application; (2) providing a tutorial on its setup and use for data collection and data transformations; and (3) outlining the benefits it provides to ABA research and practice.

Countee Features

Electronic Keys

In Countee, users can program and label keys for measuring either frequency or duration of one or more events. Throughout the article we will use the term key to refer to the icons that will appear on the screen during data collection that represent the chosen target behaviors. During data collection sessions, programmed keys appear vertically on the screen. If the all the programmed keys do not fit on the screen in a single vertical column, additional column(s) will appear. To record the occurrence of a frequency event the user simply has to tap the key. For duration, the user will have to tap the key at the onset and again at the offset of the behavior. If subcategories of main keys are also programmed, they will appear as a dropdown menu when the main key is selected.

In the application the frequency and duration keys can be programmed to represent many environmental events or stimuli in addition to behaviors of interest, such as contextual variables related to a target behavior (e.g., location, staff, noise levels), precursor behaviors, and responses as correct, incorrect, or prompted (see Table 1 for additional examples). As such, Countee can be programmed so a trained user can record almost any event of interest (see Heward et al., 2022, for examples) and any dimension of an event (LeBlanc et al., 2016). With the appropriate setup of the frequency and duration keys, the user can also easily design, calculate, or derive additional measures such as latency, interresponse time, interval-based data, and percentages. Users can also program an error key as a frequency event that can be pressed if an error of commission is made—an “error” appearing in the output file would indicate that the previous entry was a mistake. The user can then remove the erroneous entry during data cleaning.

Table 1.

Examples of Countee applications with main and qualifier key examples

Data Collection Purpose	Main Key Examples	Qualifier Key Examples
ABC Data	1. Location (D) 2. Antecedents (E) 3. Behaviors (E) 4. Consequences (E)	1. Rooms or settings such as home, school 2. Access ended; instruction placed; without attention 3. Aggression, self-injury 4. Attention provided, instruction removed/maintained, preferred item delivered
Skill acquisition	Each key is a different target behavior (F)	All keys would have the same qualifiers: correct, incorrect, prompted
Implementation integrity or task analyses	Each key (F) is a different step (e.g., model provided, feedback provided)	All keys would have the same qualifiers: yes, no, n/a or correct, incorrect, prompted
Hygiene behaviors	1. Handwashing (D) 2. Food preparation (F) 3. Coughing or sneezing (F)	1. With soap, water only, hand sanitizer 2. Cleaned surface after meat, changed knife 3. Covered mouth/nose with hand/elbow, used tissue, wearing a mask
Eating	Keys for each: acceptance (F) and refusal of a bite or drink (E)	Specific foods and beverages consumed
Preference Assessments	1. One key for each item in the assessment (E) 2. Behavior when removed (E)	1. Touched it, played with it, did or did not consume (for edibles) 2. Challenging behavior, gave up item

F = Frequency, D = Duration, E = Either

For both frequency and duration keys, qualifiers or subcategories can be added for each main key (e.g., target behavior). For example, if taking a bite of food is the main key, the user can add qualifier keys indicating the type of food eaten, or if aggression is the main key, the user can add qualifier keys indicating the specific form of the behavior (e.g., hitting, scratching). Qualifier keys can also be used to record subcategories of other events (e.g., magnitude, antecedents). During data collection, qualifier keys appear as a secondary drop-down menu that allow the user to select a subcategory. Selecting the main and subsequent qualifier key will record a single event (i.e., main key and qualifier) in the data output file (see Fig. 1). Once qualifier keys are created in Countee, one of them must be selected for an entry to be recorded during data collection. For example, if aggression (main frequency or duration key) is selected, the user must also select hitting or another designated qualifier key. Otherwise, the entry will not appear in the data output file. If the user selects a main key by accident, they can select “cancel” from the drop-down menu to exit without making a data entry. Besides the benefit of being able to take data without cluttering the main screen, the addition of qualifier keys can also reduce errors of commission as the secondary menu allows the user to cancel the initial entry.

Fig. 1.

Sample output of Countee data in Microsoft Excel

The user can also assign frequency and duration keys as being mutually exclusive, which prevents data collectors from accidentally taking data on two events that cannot occur at the same time. Although the feature is available for both types of keys, it is most applicable for duration-based events that are incompatible with other events being recorded during the observation. If keys are marked as being mutually exclusive, while one event is being recorded, other keys will be blacked out until the first key is turned off. For example, if someone is recording the duration of both flopping and in-seat behavior in a classroom, while duration of flopping is being recorded, the key to record in-seat behavior will be blacked out until flopping key is turned off.

Output

Completed sessions in Countee are stored in the application and can be sent as a CSV file via email, messenger app, or AirDrop, stored directly into Dropbox, or saved in a file on the device. The CSV file can also be opened and saved as a Microsoft Excel file. The output file includes the session name and date, the total session duration, comments or notes, the total frequency and duration of each event, and second-by-second time-stamped events in the order they were recorded (Fig. 1).

How to Set Up the Application

The first step is to create a new template for your data collection (Fig. 2). This can be done by pressing the plus sign ( +) in the top right-hand corner of the home screen of the application, which will result in a window appearing where the name of the template (e.g., hand hygiene) can be created. Multiple templates can be set up for different purposes (e.g., functional analysis, skill acquisition) and stored in the application on the device; however, templates cannot be shared across devices at this time. Session duration can be changed each time the user starts a new data collection session, and sessions can be ended before the selected time has run out.

Fig. 2.

Screenshot of Countee screen to create a new template and add or modify keys

Once the template has been created a new screen will open that will allow you to enter a default session duration for that template, a description, and add keys for each event. Electronic keys can be added by pressing “add new + ” on the center-right part of the main template screen. When “add new + ” is pressed on the template home screen, another screen will open that contains spaces to type the name of the key (e.g., the behavior or event) and a description, and select the type of measurement (i.e., duration or frequency) and a color from a dropdown menu (Fig. 3). Colors can be used to organize keys in a variety of ways to increase saliency between them and facilitate data collection for the user. For example, colors can be used to differentiate duration events from frequency events or to differentiate behaviors to decrease from those to increase. Qualifier keys for each duration and frequency event can be added on the same screen by selecting “add new + ,” also on the center-right part of the screen. Keys can also be added or modified on a preexisting template by opening the template and pressing edit at the top-right corner of the screen. To avoid losing changes made to templates or keys, the user must hit save at the top right corner of the screen before pressing the back arrow to go to the previous screen or closing the application.

Fig. 3.

Screenshot of Countee depicting the screen to program keys, add qualifier keys (left) and how they appear during data collection sessions (right)

On the same screen where details about a given key are entered (e.g., name, data type, color), all other previously programmed keys will be listed on the bottom portion of the screen (Fig. 4). The default is that all keys will be active at the same time. Therefore, any mutually exclusive events will need to be deactivated. When programming the event key at the top of the screen, shifting the toggle to the left on the mutually exclusive event key below will restrict the ability to select that key during the data collection session. Note that the process of deactivating keys while a specific key is running needs to be repeated for each key. Referencing Fig. 4 as an example, shifting the toggle to the left for behavior 3 deactivates the ability to select it when behavior 2 has been selected and is active (i.e., turned on) during data collection. However, doing so will not automatically deactivate behavior 2 when behavior 3 is active. To deactivate behavior 2, the user must repeat the above process when editing behavior 3.

Fig. 4.

Screenshot of Countee depicting how to program mutually exclusive events (left) and how the keys appear during data collection (right)

To run a session, select the appropriate template from the Countee home screen and touch the “new session” button on the top-right. Fill in the name of the session in the window (e.g., date, participant ID number) and press “create.” A new screen will open, showing the keys that have been created. To begin taking data, press the “start session” button. To end or pause the session before the time runs out, press the “pause” button and select the option to either resume the session again or end it. At the end of a session, a screen appears that allows the user to change the session name and enter any notes that they may want to appear in the CSV or Microsoft Excel output file. To transfer the data to a computer or a cloud service (e.g., Dropbox) or to share it via email, press “my sessions” at the bottom right of the home screen, then the white curved arrow at the top. (Fig. 5). Sharing only specific sessions can be done by touching the boxes to the right of those sessions, or all of the sessions can be shared by pressing “select all” at the bottom-middle part of the screen. Once the files for sharing have been selected, pressing “share” on the bottom right part of the screen will produce a window from which the modality to share (e.g., email, Dropbox) will appear. Files will be stored on Countee until manually deleted.

Fig. 5.

A screenshot depicting the file sharing screen

Calculating Derived Measures

With appropriately programmed duration and frequency keys and Countee timestamps of each event, it possible for users to calculate percentages, latency, interresponse time, and discontinuous measures (e.g., partial interval) quickly and accurately. Although the raw data are visible in the Countee by opening the session in the application, most of these calculations will be more easily performed once the raw data have been transferred to an Excel or CSV file. If the user wants to calculate the percentage of sessions in which an event has occurred (e.g., specific behavior), they could simply pull the percentages from the summary at the top (see Fig. 1). If the user wants to calculate the percentage of one event in proportion to another, again, they could take the totals of each event from the summary at the top and divide one by the other.

How one calculates latency from one event to another can be done in two different ways, depending on how the keys are programmed (Fig. 6). For the first, the user would have to program a frequency key to represent the presentation of an event (e.g., instruction presented) and another key to represent the other event (e.g., follows instruction). The user could then calculate latency by subtracting the time of the second event from the time of the original event. For example, if the instruction occurred at 10:03 of the session and following the instruction occurred at 10:08, the latency would be 5 s. The other option is to create a duration key to represent the first and second events. The key would be turned on when the initial event occurs and turned off when the second event occurs, resulting in the latency being calculated automatically in the output file. The average latency can be calculated by dividing the total latency (marked as sum in the output) by the number of occurrences.

Fig. 6.

Calculating latency with two keys (left) and with a single duration key (right) from the output data in Countee

Interresponse time can be recorded with frequency keys in the same way as latency, except only one key would be required per target response, which would be selected each time the event occurred. Calculating interresponse time is almost identical to calculating latency; however, instead of taking the difference between two keys, interresponse time involves calculating the difference between selections of the same key. The use of a duration key for measuring and calculating interresponse time is not recommended because it may not be feasible to turn on and off the key repeatedly as events occur and requires additional calculations of time between duration events.

If there are barriers to collecting continuous data (e.g., high frequency, user attention), discontinuous data can be collected by programming frequency keys to denote intervals for when the event occurred or did not occur. This would be similar to taking interval data on paper, except that instead of indicating that an event did or did not occur in an interval by checking a box, the data collector would be pressing a key. An advantage of using Countee rather than paper is that it could be easier to calculate percentages of intervals directly from the output file as the number of intervals in which the event occurred (“intervals marked”) and the number of intervals in which the event did not occur (“intervals unmarked”) are summarized at the top. The user could then calculate the percentage of intervals in which the event occurred (occurrence measure) as well as the percentage of intervals in which the event did not occur (nonoccurrence measure) by dividing either number by the total number of intervals (sum of intervals in which the event occurred and intervals in which the event did not occur).

Calculating Interobserver Agreement

If session data is going to be used to calculate interobserver agreement (IOA), it is essential that data collectors begin their sessions at the exact same time to ensure that timestamps on the two output files match with events that occurred in real-time. For live observations, one can determine if the sessions began at the same time by looking at the time at the top of the output file (see Fig. 1). For video observations, it might be important to ensure data collection begins at a specific time in the video. Even though data collectors may communicate with each other regarding session start times, observations should still be completely independent.

Proportional agreement can be assessed by uploading CSV files onto the Countee website; all other types of IOA can be calculated directly using the data in the output files in Microsoft Excel. To calculate proportional IOA on the Countee website (https://www.counteeapp.com/ioa/), the user has to upload the CSV file for each observation, enter in the interval duration (e.g., 10 s), and press send. A window will then appear, to allow the user to save the file with the IOA percentages on their computer. To calculate total IOA using Microsoft Excel files, the user could use the data in the summary portion at the top. All other types of IOA can be calculated using the timestamped data on the bottom portion of the output in Microsoft Excel (see Reed & Azulay, 2011, for a tutorial on calculating IOA).

Discussion

Benefits for Research and Practice

There are several benefits to using Countee in research and practice that require direct observation, which are tied to the collection of the data itself. The first is that it could reduce the amount of needed equipment, such as counters, timers, clipboards, printed data sheets, and writing instruments. In addition, it is commonplace for people to carry a phone, therefore using one for data collection is often inconspicuous in comparison to other data collection methods. Due to the running timer which timestamps individual events in sequence during the observation as well as the date and time of the observation, it becomes more difficult to falsify data, which is a documented concern. In Morris et al. (2022), 76% of survey respondents reported that they doubted the accuracy of collected data and that 68% reported that data collectors were filling in data after the event. In cases such as these, if data falsification were in question, relevant parties could confirm the date, time, and duration of the observation with the timestamps in the output file. Although data could be falsified using the Countee application, the user would have to keep the application running for the entire intended session length and make entries intermittently through that time period to mimic a how real data would appear. It would be less effortful and time consuming to falsify data on pen and paper because all of the “data” could be recorded in the time required to write it down.

Although a complete description of HIPAA compliance is beyond the scope of this article, it is important to note that Countee has features that support HIPAA compliance and participant confidentiality. First, because files can be named by the user, participant identification numbers can be entered rather than names, and all personal health information can be excluded. Second, all tablets and phones can be password protected, which adds additional security in the case of loss or theft. In addition, lost or stolen devices can often be located through global positioning systems and the stored information can be deleted remotely if the device cannot be recovered. Finally, once files are uploaded to a secure location (e.g., cloud storage), they can be deleted from the application.

Another benefit is that it is easy to upload or save data files to a computer or cloud storage (e.g., Dropbox) with Wi-Fi or cellular data. Those files can then be opened and saved as a Microsoft Excel file, uploaded directly into data analysis programs such as R (R Core Team, 2023), or copied and pasted into graphing software such as GraphPad Prism. This reduces the time and potential errors related to manually entering data from data sheets to a computer file. Files can also be easily shared with collaborators in their raw form without having to have someone enter it into data software. It also allows for raw data to be uploaded onto open databases, which is requirement for publication in some journals and for many grant funders. The use of open databases can increase the authenticity of research findings and reduce the risk of falsified data. Finally, raw data in the form of a CSV or Microsoft Excel file is that it allows for multiple forms of data transformation with the direct use of your file.

The ease of use of Countee and transfer of data between parties also hold several advantages for clinical practice. First, even clinics or centers that have extensive data management systems in place (e.g., CentralReach) could use Countee as a supplement for programs that require mobility, a less conspicuous data collection method, or more sensitive measures (e.g., second-by-second data or subcategories). For example, during functional analyses, toileting routines, hand hygiene, and feeding sessions. In addition, for data collected outside of clinics and centers (e.g., field trips), the use of a phone over a clipboard or computer or tablet-based system would reduce attention from outside observers, protecting client privacy. Countee is also a feasible option for caregivers, educators, and others to collect data—not only is it easily accessible, but it also allows for immediate data transfer to supervising clinicians without requiring the scanning or emailing of documents. To avoid having practitioners or researchers access a caregiver’s or educator’s phone, behavior analysts could guide a caregiver through the steps verbally with corresponding screenshots of the Countee application. This article might serve as a guide for creating tutorials that include programming keys, qualifiers, etc. that are unique that client.

Directions for Future Research

Outside of the 37 known manuscripts that have reported using Countee in their research it is relatively unknown the extent to which the application is being used in research and practice. A survey could be conducted to determine how many people are using it and how. In addition, although the advantages of electronic data collection systems over pen and paper have been documented in previous research (e.g., Tapp et al., 2006), future research could compare Countee with pen and paper as well as other electronic data collection systems. In particular, in addition to reducing the time required for data transfer and analysis, one could evaluate if its use also increases accuracy and reduces errors (e.g., with the use of mutually exclusive events) or facilitates data collection on multiple behaviors and subcategories at the same time. In follow-up to the survey conducted by Morris et al. (2022) on data integrity, research could be conducted to determine if the use of Countee (with timestamps) decreases the frequency of data being collected at the end of the session or well after the behavior has occurred. Additional research could also be conducted on how to empirically teach individuals to select an appropriate data collection method as well as how to best train them on that method.

Limitations

Although the Countee application carries a number of benefits, there are some limitations that are worth noting. The first is that it is not possible at this time to share templates across devices, so templates must be set up on each new device and with each new user. Another is that it is not appropriate for managing data collection for multiple skill acquisition programs across multiple learners in a clinic or school. It may be useful, however, for a subset of behaviors in those programs that require more detailed data collection, such as feeding programs and assessments. Although there are methods for managing errors of commission that were discussed in this article, once a key is pressed and entered into the record, the only way to remove it is by taking it out of the final output file. Finally, the notes section is only available at the end of the session, so any notes that are required during the session would have to be taken using another method.

Conclusion

Countee is becoming a more widely used electronic data collection system with many applications and benefits. It has several features that allow for data to be collected on frequency and duration of behavior as well as other derived measures with built-in protections against errors of commission (e.g., mutually exclusive keys). The use of Countee is a secure data-collection system that may improve the rigor of data being collected through direct observation by deterring data falsification and facilitating data transfer and sharing and the assessment of (e.g., IOA assessment). Data files can be easily uploaded to cloud storage, transfer errors are reduced, security is enhanced, and sharing of files is simpler than it would be with paper and pencil methods. With these benefits, the dissemination of Countee might encourage ease of access and facilitate the use of direct measurement research in behavior analysis and other fields.

Data Availability

The article has no associated data.

Declarations

Conflicts of Interest

The authors have no conflicts of interest to disclose.