83
observers available, an adequate sampling method must be chosen. Among some
authors (Altmann 1974; Martin and Bateson 1993; Bart et al. 1998; Mann 1999,
2000) there are differences in naming the methods, some divergences in definitions,
but all of them discuss important considerations about the sampling methods, arising questions, restraints and problems about bias, over and underestimations, randomness, dependency, pseudo-replications, etc. It is crucial to understand all
methods, assumptions and statistical consequences before defining the research protocol. A summary of the most frequent and common sampling methods used in LBS
studies available in the literature is presented. However, for the complete information and assessment of the methods, search the above-mentioned references. Before
starting, it is important to decide on systematic rules: which subjects to watch and
when, how the behaviors will be recorded (Martin and Bateson 1993) and for how
long (Mann 1999).
“Ad libitum sampling”: there are no systematic constraints when the observer simply notes
down whatever is visible and what seems relevant at the time (Martin and Bateson 1993).
Altmann (1974) describes it as “field notes”. This method will be clearly biased towards
more conspicuous behaviors and data are not appropriate to estimate behaviors rates. It may
be used during preliminary observations and pilot studies while creating the ethogram and
delineating research questions (Mann 1999).
“Focal sampling” or “focal-follow”: one individual or two (or some other focal unit
within an adequate size for reliable data collection) are followed. The focal unit must have
specific characteristics, or must behave in a way, which permit its identification and track.
The unit will be observed for a specific amount of time, until it leaves the boundaries of the
study area or until viewing conditions from land station gets rough due to glare, haze, or
higher sea state. There should be rules in the observation protocol for deciding what have to
be done when the focal unit is out of sight or “disappear” (Martin and Bateson 1993).
Longer focal sampling observations are encouraged in order to avoid rare behaviors underestimation (Mann 1999), although observers fatigue should be taken into account in this
context (Senigaglia 2011). The “focal continuous sampling” is when all behaviors performed by the focal unit are recorded with the exactly time of occurrence and duration,
enabling to reproduce all that was observed (Martin and Bateson 1993). Continuous data
are the richest source of information on social behavior, including information on details,
sequences, rates and durations of behavior for individual animals (Mann 1999). This
method is very demanding for observers and reliability of the data can be compromised
when animals are active, or the focal unit is too large (Mann 1999). Altmann (1974) recommends this method for just one, or a maximum of two individuals.
“Focal-group sampling”: a group (more than one individual) is followed; however, as
already mention above, to observe and record all behaviors of all members of a group is
very challenging leading to biased data. The “focal-group continuous sampling” is practically impossible without the use of video recording (Mann 1999).
“Instantaneous or point sampling”: the focal unit is sampled periodically on determined
fixed moments, i.e. the behavior is recorded only if it occurs at the exactly moment of the
sample point. As aquatic mammals dive between surface periods, they might be unavailable
during some data collection points. The “one-zero or timing sampling” is when the focal
unit is sampled periodically on fixed intervals and the observer records whether or not a
behavior had occurred (Martin and Bateson 1993). Both, instantaneous sampling and onezero sampling, do not give true behavior frequencies nor durations (e.g., Altmann 1974;
Mann 1999), so their use must be specific to certain research goals where this kind of information is not necessary.
“Scan sampling”: each individual, is observed for a short period and their behavior is
recorded, but not in a continuous manner. There is no predetermined time for a scan, it can
be conducted at regular intervals, or as quick as possible (Mann 1999). It may last few
4 Land-Based Station Studies of Aquatic Mammals in Latin America: Understanding…
observers available, an adequate sampling method must be chosen. Among some
authors (Altmann 1974; Martin and Bateson 1993; Bart et al. 1998; Mann 1999,
2000) there are differences in naming the methods, some divergences in definitions,
but all of them discuss important considerations about the sampling methods, arising questions, restraints and problems about bias, over and underestimations, randomness, dependency, pseudo-replications, etc. It is crucial to understand all
methods, assumptions and statistical consequences before defining the research protocol. A summary of the most frequent and common sampling methods used in LBS
studies available in the literature is presented. However, for the complete information and assessment of the methods, search the above-mentioned references. Before
starting, it is important to decide on systematic rules: which subjects to watch and
when, how the behaviors will be recorded (Martin and Bateson 1993) and for how
long (Mann 1999).
“Ad libitum sampling”: there are no systematic constraints when the observer simply notes
down whatever is visible and what seems relevant at the time (Martin and Bateson 1993).
Altmann (1974) describes it as “field notes”. This method will be clearly biased towards
more conspicuous behaviors and data are not appropriate to estimate behaviors rates. It may
be used during preliminary observations and pilot studies while creating the ethogram and
delineating research questions (Mann 1999).
“Focal sampling” or “focal-follow”: one individual or two (or some other focal unit
within an adequate size for reliable data collection) are followed. The focal unit must have
specific characteristics, or must behave in a way, which permit its identification and track.
The unit will be observed for a specific amount of time, until it leaves the boundaries of the
study area or until viewing conditions from land station gets rough due to glare, haze, or
higher sea state. There should be rules in the observation protocol for deciding what have to
be done when the focal unit is out of sight or “disappear” (Martin and Bateson 1993).
Longer focal sampling observations are encouraged in order to avoid rare behaviors underestimation (Mann 1999), although observers fatigue should be taken into account in this
context (Senigaglia 2011). The “focal continuous sampling” is when all behaviors performed by the focal unit are recorded with the exactly time of occurrence and duration,
enabling to reproduce all that was observed (Martin and Bateson 1993). Continuous data
are the richest source of information on social behavior, including information on details,
sequences, rates and durations of behavior for individual animals (Mann 1999). This
method is very demanding for observers and reliability of the data can be compromised
when animals are active, or the focal unit is too large (Mann 1999). Altmann (1974) recommends this method for just one, or a maximum of two individuals.
“Focal-group sampling”: a group (more than one individual) is followed; however, as
already mention above, to observe and record all behaviors of all members of a group is
very challenging leading to biased data. The “focal-group continuous sampling” is practically impossible without the use of video recording (Mann 1999).
“Instantaneous or point sampling”: the focal unit is sampled periodically on determined
fixed moments, i.e. the behavior is recorded only if it occurs at the exactly moment of the
sample point. As aquatic mammals dive between surface periods, they might be unavailable
during some data collection points. The “one-zero or timing sampling” is when the focal
unit is sampled periodically on fixed intervals and the observer records whether or not a
behavior had occurred (Martin and Bateson 1993). Both, instantaneous sampling and onezero sampling, do not give true behavior frequencies nor durations (e.g., Altmann 1974;
Mann 1999), so their use must be specific to certain research goals where this kind of information is not necessary.
“Scan sampling”: each individual, is observed for a short period and their behavior is
recorded, but not in a continuous manner. There is no predetermined time for a scan, it can
be conducted at regular intervals, or as quick as possible (Mann 1999). It may last few
4 Land-Based Station Studies of Aquatic Mammals in Latin America: Understanding…