Basic research of complex behaviors has led to a dramatic increase in the use of rodent behavior tests (1). Until recently, these were videotaped and laboriously scored by hand. Automated video tracking systems are evolving in sophistication but are costly, require dedicated space/computing resources, and are ill-suited for infrequent, exploratory studies or scoring of behaviors not amenable to computer vision. Hence, traditional manual scoring remains important for many laboratories. In using the elevated plus maze (EPM), a commonly employed measure of anxiety (2), several drawbacks were noted in manual scoring: (i) there is a steep learning curve as the scorer needs to negotiate several stopwatches and mechanical event counters simultaneously; (ii) because of the complexity and number of events to monitor, more sophisticated analyses may need to be conducted on a second pass or by multiple scorers; (iii) errors occur in transcribing the manually acquired data to spreadsheets for statistical analyses; and (iv) scorer fatigue contributes to decreased throughput. We reasoned that computer-assisted scoring would be simpler, more intuitive, less tiring, and less error prone, but could not find such a tool in the public domain.

We report here the development of a program to facilitate EPM scoring. EPMscore.exe is freely available as a standalone DOS executable program (www.mbni.med.umich.edu/mbni/files/EPMscore.zip) that runs on Windows PC and produces a universal tab delimited output. The EPM test is videotaped as usual (2). During subsequent scoring, the user tracks arm entries on a topographically mapped layout of the number keypad with the right hand (Figure 1) and records specific behaviors—partial (f)alls, (r)ecoveries, head (d)ips, (s)tretched attend postures, and (v)ertical explorations—with the left hand in standard QWERTY position. An entry is scored when all four paws enter an arm. The program extracts core measures: percent entries into the open arm, percent time spent in the open arm, and number of closed arm entries (2,3), but also reports many other potentially useful measures (Table 1) that may have ethological validity (4,5). Since the user is free from tracking timers, more focus is available for simultaneously scoring behaviors that would otherwise be scored in a separate pass or only after considerable experience. The interface is intuitive, and latency to scoring competence is shorter. EPM scoring of many subjects in tandem is tiring and prone to omission and commission errors. Users consistently report the gaming quality of EPMscore renders the task less fatiguing. The data may be imported directly into a spreadsheet, thereby eliminating potential transcription errors. Collated summary data facilitates sorting for final statistical analysis. In testing trials, results obtained by a less experienced scorer using EPMscore agree with measures acquired by manual scoring (Figure 2).

Figure 1. Keystrokes used to score activity on the elevated plus maze. (Click to enlarge)

Figure 2. Comparison of measures acquired manually versus with EPMscore. (Click to enlarge)

Table 1. Measures Produced by EPMscore and Corresponding Column Headers in the Output File (Click to enlarge)

EPMscore permits richer mining of behavior results. Since every event is recorded with either a time or interval stamp, it is easy to ask questions not feasible using traditional hand scoring (e.g., whether the frequency of transitions, head dips, etc., vary over the duration of the test, or whether specific behaviors occur more frequently in a particular arm of the maze). Temporal coding also captures unusual behaviors for later analysis. For example, latency to open arm entry may be misleadingly short for the occasional subject that darts into an open arm to escape the handler, only to quickly enter a closed arm and subsequently behave similar to others in the cohort. EPMscore permits distinction of this pattern by reporting latency to open arm entry after a user-specified delay. We have not formally tested the validity of the latter measure, but it highlights the type of alternate measure that is easily gleaned without additional user effort and may improve behavioral scoring.

EPMscore is moderately flexible and provides some degree of error correction. There are inconsistencies in the literature as to whether center time is scored. Hence, it is tracked in EPMscore at user discretion with a user-definable minimum duration to avoid coding for subjects that pass quickly from one arm to another. Likewise, there is a user-definable minimum time before accepting an arm entry, thereby eliminating anticipatory miscoding by the scorer. Each keystroke is recorded and saved as raw data, but transitions less than the minimum time are ignored in the corrected summary data. If desired, minimum time may be set to zero to force all data points into the analysis. With respect to transitions, EPMscore reports separate measures for homologous (open-to-open or closed-to-closed) versus heterologous (open-to-closed or closed-to-open) arm entries. The latter are considered by some to be true open or closed entries, but the total number of transitions is also a measure of locomotor activity (4), and investigators may choose to consolidate these measures during analysis. Again, the core measures are available, but EPMscore provides details for downstream mining. The application also permits an interesting option not typically considered. Subjects occasionally fall off the open arms of the maze. Homemade and commercial mazes vary in design, and some mazes have a lip or mesh cliff that permit the subject to catch their fall. These partial falls may be noted, but the time spent in struggling to regain perch can be substantial and is usually subsumed in the open arm time. In EPMscore, the fall time is easily captured and subtracted to report an optional corrected open arm time.

In summary, EPMscore is a computer-based event recorder that has a number of desirable features. It is very easy to adopt, and user attention is subjectively improved for an otherwise mundane task. Data transcription errors are eliminated. We anticipate that with minimal training, EPMscore will permit scoring of subjects directly during testing—a potentially significant time savings. The core algorithm in EPMscore may be employed for other behavior tests such as the elevated zero maze (6), light-dark box (7), and social interaction test (8). Computer-assisted scoring is particularly useful when temporal coding is required (e.g., integration of behaviors over 5-s intervals in the forced swim test) (9). Programs like EPMscore have the potential to increase throughput and reliability of manually acquired behavior data.