Definitions.
Developing robust, measurable definitions is a crucial step of the research process. There are a few key steps we take when beginning to develop definitions:
1. Observe the animal(s) before narrowing down variables of interest (behaviors, wound characteristics, etc.). Collect a variety of example videos and photos to help you make decisions before your experiment begins.
2. Consider how a particular behavior or parameter connects to and addresses your research question.
3. Review existing literature and definitions developed by others.
4. Develop robust, measurable definitions. Try using them to score the outcomes with direct observation, pictures or videos of your study animals. Seek discussion with others about your definitions and revise based on the feedback.
5. Continue collecting video or photo examples that demonstrate your definitions. You'll need a lot of examples like this when you begin reliability training. Collecting some of them at this stage will helpful in the long run.
2. Consider how a particular behavior or parameter connects to and addresses your research question.
3. Review existing literature and definitions developed by others.
4. Develop robust, measurable definitions. Try using them to score the outcomes with direct observation, pictures or videos of your study animals. Seek discussion with others about your definitions and revise based on the feedback.
5. Continue collecting video or photo examples that demonstrate your definitions. You'll need a lot of examples like this when you begin reliability training. Collecting some of them at this stage will helpful in the long run.
When we develop definitions (Step 4, above), we typically ask ourselves a few key questions:
1. Is it operational? Operational definitions provide clear direction about how to score events or characteristics, and do not rely on an assumption that observers will classify these in the same way.
|
This video shows dairy calves outside. The one with the ear tag labeled 5469 is performing tongue flicks. Video credit: Blair Downey
|
Bad example:
Tongue flicks = The animal extends its tongue repetitively Better example: Tongue flicks = The animal extends its tongue out of its mouth at least 3x with < 2 s between tongue extensions. The tongue does not touch anything while extended. Note that these definitions describe a "bout", or a cluster of movements occurring in quick succession, rather than individual occurrences of each tongue extension. For more considerations specific to bout definitions, see question #5, below. |
2. Does it create circularity? Circular definitions, or those that use the variable of interest in the definition, involve assumptions that may not be shared amongst all observers.
|
This video shows 2 cats lying next to each other. The cat on the right is grooming itself. Video credit: Blair Downey
|
Bad example:
Grooming = The animal is grooming itself Better example: Grooming = The animal is touching hair on its own body with its tongue or mouth while the head moves in the same direction relative to its tongue or mouth movement. |
3. Does it make assumptions about an animal's intentions? Animal intentions are not known to us, and thus cannot be measured. Using intention in a definition involves assumptions about what that looks like that may not be shared amongst observers.
|
This video shows a beef steer (castrated male cattle) from the side touching a wooden gate with his nose. Behind the wooden gate is a blue feed bin. If the wooden gate was pushed, it would have lifted up out of the way. Since the gate does not open (lift up) after the steer made contact with his nose, this would be a push attempt. Video credit: Rachael Coon
|
Bad example:
Push attempt = The animal tries to open the gate but fails Better example: Push attempt = The animal makes contact with the gate with part of their own body but the gate does not open |
4. If it requires an estimate of distance or size, can that be determined in reference to elements in the frame of view? Distance and size are difficult to approximate, and may change with the perspective of the observer. Using a reference point that is visible from all positions prevents guessing. With size, using a reference point that is either in the frame, or one that is typically well-known to a general audience, can also help.
This photo is an overhead view of a cat walking towards a seated human. The first dotted line represents 0.5 meters away from the human. Given our definition, this cat would be considered "in proximity" at this moment, as its body has crossed the first dotted line. Photo credit: Carly Moody
|
Bad example:
Proximity = The animal is within 0.5 meters of the human. Better example: Proximity = The animal is within 0.5 meters of the human, as indicated by, at minimum, the animal’s head crossing the first dotted line closest to the chair. Note that in some experimental settings, it isn't possible to mark distances due to camera and observer angles or environmental features. For example, if observing animals in a straw-bedded pen, it might not be possible to paint a distance line on the straw, as the animals could disturb that line. In these cases, we use reference points that are clearly visible from all observer or video positions, such as a head-length or body-length. In this case, we might instead say "Proximity = the animal's nose is within one head-length of the human." |
5. If relevant, are the criteria for ending a bout detailed? If the end of a bout is not described, we are relying on a shared assumption about when to stop scoring the behavior.
|
This video shows a dairy calf performing a head shake. Video credit: Alycia Drwencke
|
Bad example:
Head shake = The animal rotates its head at least once to both sides (left and right) in rapid succession. Better example: Head shake = The animal rotates its head at least once to both sides (left and right) in rapid succession. To count as a new head shake, the head must return to a level position and remain still for at least 2 seconds |
Note:
While we think about definitions (and their flaws) when we measure behavior, the aforementioned criteria also apply to definitions for other variables, like wound types, or health scoring categories. For example, "crust," a type of wound tissue that can occur after ear tagging, could be defined as "raised dried tissue of any size that is tan, yellow, red, black, or pink." This definition would be more robust than one classifying crust as "crusted tissue" (circular) or "small dried skin" (not operational). Similarly, if we wanted to consider size of a wound, using a definition that said "Any open wound or swelling greater than 1" or 2.5 cm (the size of a quarter) in height" would be better than saying "a small wound".
