Online Zoologists

Event Recorder Information System (ERIS)

Last updated: 960822

EVENT RECORDING WITH MULTIPLE OBSERVERS: LESSONS LEARNED

Wesley R. Elsberry & Diane J. Blackwood (Marine Bioacoustic Laboratory, TAMUG, 5007 Ave. U, Galveston, TX 77553 USA)

(Presented at the 1996 Conference of the Animal Behavior Society, Flagstaff, Arizona, 4 August 1996.)

Contents

ABSTRACT

The need to record concurrent behaviors from multiple focal animal follows and focal animal follows combined with human interactions drove the development of a an event recording information system (ERIS). The capabilities of this software system include simultaneous recording on two platforms with a serial link, the configuration of user-defined tokens both from menus and from a file, and the specification of a time base for recording. The system was used for analysis of acoustic recordings of lekking male Greater Prairie Chickens (Tympanuchus cupido pinnatus) and videotapes of a stranded male Atlantic Spotted Dolphin (Stenella frontalis). Recording prairie chicken vocalizations required the use of user-defined tokens for quick response. Recording the correlation of dolphin behavior with human movement patterns required the capacity for multiple observer use. The use of user-defined tokens brings the issue of lexical analysis into specification of behavioral codes. A hierarchical approach to naming behavioral codes was used for the dolphin behavior project. The ERIS software is freely available.


Event Recording: What It Is


Event Recording: Determining Suitability of the Technique


Event Recorder Information System: A Dozen Lessons Given by the Data

Lesson 1: Choose the Right Platform

The Event Recorder Information System (ERIS) provides event recording capability for IBM PC compatible machines running MS-DOS or equivalent operating systems. The first event recorder that we wrote was done in BASIC on a TRS-80 Model 100 portable computer. There are not many of these computers available now, though, so moving to the IBM PC platform opened up many more available machines and options.

Lesson 2: Use Tokens

The input paradigm used is that of tokens , which are strings of characters which are recognized by the ERIS program. With no tokens defined, the program accepts input of short strings terminated by pressing the Enter key. When tokens are defined, the user input is scanned as each character is entered, and a match causes the recognized token to be queued with its timestamp. Reason: Always terminating entry with some keystroke is inefficient when events may follow one another in close temporal sequence. Being able to specify a hotkey -like response to certain keystrokes allows for better resolution of fast-paced event sequences.

Lesson 3: Queued Entry

ERIS has a single-entry queue, allowing the latest entry to be deleted before logging to disk. Reason: Errors in entry by observers are inevitable. Many errors are immediately apparent. Allowing erasure of the latest entry covers most error conditions. Other errors can be noted via comments.

Lesson 4: Comments

The token-recognition system can be bypassed for the entry of user comments. Reason: Especially in the case of single-character tokens, entry of information not covered in the current set of defined tokens requires that the user be able to bypass the token recognition system, since otherwise the initial character of a comment might happen to be recognized as a token and stored separately.

Lesson 5: Tokens-Only Entry Mode

When a complete and consistent set of tokens is settled upon, the ERIS system can be put into a token-only entry mode, where any user input that does not lead toward a known token is simply ignored. Reason: When a large number of tokens are being entered by the user, a common error involves simply hitting the wrong key. The user s attention needs to stay on the animal. By simply ignoring any keypress which does not lead toward completion of a token, a whole class of erroneous data inputs is eliminated.

Lesson 6: Early Binding of Timestamp

The first key-press of a token or comment causes a time-stamp to be bound to that entry. Reason: For multiple-character tokens or comments, the time a user takes to enter the token or comment is a source of potential bias, especially between common and uncommon tokens. Some event recording programs use late-binding of timestamp, such that only at the completion of the user input is the timestamp generated. This means that the difference between timestamps in late-binding is composed of the difference in time of the observed behaviors, the user reaction time, and the user token completion time. Early-binding of the timestamp eliminates the token completion time source of bias.

Lesson 7: Time Specification

The user can specify an initial time. Reason: When recording behaviors from audio or video sources, it can be helpful to match the time recorded by the ERIS to that noted on the audio or video source. Events noted on the computer can be easily returned to on the audio or video source. By default, ERIS uses the system time.

Lesson 8: Multiple Platform Use

ERIS allows two PC compatible machines running ERIS to be linked via a null-modem serial cable. Time synchronization between the machines is accomplished via a master/slave arrangement. This arrangement allows for multiple observers to simultaneously record event information. Reason: In certain circumstances, hypotheses of behavior may require the entry of data either at higher rates or of a broader range of events than can be accommodated by one observer. Being able to distribute the burden of event recording across two observers expands the range of hypotheses which can be explored via event recording.

Lesson 9: Audible Feedback

By default, ERIS emits a tone at the completion of each token. Reason: When observing behaviors, most of the time the user must commit his visual attention to the animal being observed. An audible tone serves as confirmation that token entry either is or is not proceeding without error, without requiring the user to divide visual attention between the subject and the computer.

Lesson 10: Use Hierarchical Naming of Tokens

The creation of a list of tokens was done in a hierarchical manner, where the first character defines a class or category of codes, and second or further characters further delineate until the individual token is defined. Reason: Larger numbers of tokens require more effort of recall on the part of the user. Reduction of recall for the first character helps reduce user reaction time, which when coupled with early-binding of timestamp helps reduce bias between common and uncommon token entry.

Lesson 11: Split Token Entry by Attention

The distribution of tokens entered by multiple simultaneous users should be done by consideration of what each must pay attention to. In the case of Charlie, one observer was tasked with most of the tokens that related to position or action of Charlie, and the other observer was tasked with tokens dealing with actions of humans and other environmental phenomena.

Lesson 12: Don t Trust One-Point Time Synchronization

We found that there was drift between the system clocks on separate platforms, so that even when initial time synchronization was accomplished, there could be a drift in timestamps across the two linked machines. Fortunately, our data collection for Charlie had certain events that were recorded by both observers, allowing for post-entry recalibration of timescale. The ERIS system will incorporate a periodic (30s to 1 m) resynchronization of time from master to slave in future versions.

ERIS In Use: Greater Prairie Chickens

We used ERIS in making an analysis of vocalizations of lekking male Greater Prairie Chickens (GPC) (Tympanuchus cupido pinnatus Brewster). We worked from acoustic recordings made while two males were actively vocalizing, with occasional vocalizations from a third male. Tokens were defined for start and end of booms and cackles, and another for pwoiks. The time course of pwoiks was too brief to accord it a duration.

The raw timestamp/token data file was processed using a custom Perl script to extract boom and cackle durations and frequencies of interruptions. This data was imported into an Excel spreadsheet, where chi square and Freeman-Tukey deviate statistics were calculated.


ERIS In Use: Charlie, An Atlantic Spotted Dolphin

We used ERIS for analyzing the movement patterns and association of vocalizations with time for a rehabilitating stranded Atlantic spotted dolphin (Stenella frontalis). The source was videotape taken from a monitor camera that was suspended above the recovery pool, where acoustic information from a hydrophone in the pool was recorded on an audio track of the videotape. Two observers watched the video and recorded simultaneously.

There were a wide variety of events and behaviors that we wished to record. An initial list of codes was generated. That list proved too difficult for practical use, as each token was arbitrarily named. A second list of codes was generated, this time using a hierarchical approach to naming. All vocalization codes had an initial 'S' (sound) character, all movement codes had an initial 'K' (kinetic) character, and so forth. The advantages of using a hierarchical naming system follow from the ease of remembering the complete set of codes coupled with the early-binding of timestamps. The user's first keypress for any token need only classify the broad code type, and further time for completion of the token does not influence the timestamp value.

Again, the raw timestamp/token data was processed using a custom Perl script to produce output suitable for import into an Excel spreadsheet.


REFERENCES

Related links