Fig 1.
Schematic of the Automated Human Scent Olfactometer (AHSO).
Table 1.
Search and Rescue (SAR) team information.
Teams with the same superscript share the same handler.
Fig 2.
Positive control test barrel search setting.
Four barrels were placed in an open room. A 4 cm hole was drilled on each lid. Binders on the side of the barrels prevented the barrels from rolling.
Table 2.
Description of the testing odors.
Fig 3.
Schematic of the training and testing procedure.
During testing one of two volunteers were in the chamber and dogs were trained to whole human scent. The same two volunteers served as the source for the testing stimuli. During both positive controls dogs received 5 testing trials with the volunteer in the chamber as in training. During the Human No-Breath test the volunteer was in the chamber with a snorkel and a scuba mask. The breath of the volunteer was exhausted from the chamber through a PVC pipe attached to the side of the chamber during testing. A pump carried the headspace of the volunteer without breath to an olfactometer port. During the breath test a volunteer exhaled in a vial and a vacuum pump carried the breath to the olfactometer port. The distractor control consisted of presenting dogs with a novel random odor (Isobutyl propionate). Dogs received a training session in between testing sessions to ensure they still were responsive to the whole human scent target.
Fig 4.
Vertical lines indicate the initiation of the different training Phases. The color of the data point indicates the volunteer that served as the human scent (HS) source. Dashed line indicates 0.85 proportion of correct responses, our training criterion. Dogs completed their training within 28 sessions. At the end of the experiment, the proportion of correct responses for both dogs was> 0.90.
Fig 5.
Data points are the mean ± 95% confidence intervals. Introducing novel distractor odors increased the proportion of false alerts during the first two sessions. After the second session with novel distractors the proportion of false alerts reduced and by the end of the phase dogs did not show any false alerts to the distractors. The distribution of false alerts by distractor odor shows more responses to the person hygiene (PH) products of the volunteers and to the control chamber than to the other distractors.
Table 3.
Performance of SAR teams (N = 5) in the AHSO and the Control test.
Fig 6.
Back transformed LSmeans ± 95% confidence intervals of the amount of time (s) dogs spent sniffing the target odor (Human scent) and the different distractors during a trial.
The main effect of odor was not statistically significant (P = 0.24). However, on average, dogs sniffed more time to the human scent than all other distractors.
Fig 7.
Dogs (N = 6) average probability of alert and sniff time ± 95% confidence interval to the different testing odors.
Data points with different superscripts within a panel are statistically different from each other. A) shows dogs response rate to the different testing odors. B) Shows dogs sniff time (s) to the target and testing odors. Dashed line indicates the 3 s mark which was our alert criterion.
Fig 8.
Averaged probability of alert to the different testing odors during a trial ± 95% confident interval.
Dashed line indicates the different sessions.
Fig 9.
Dogs (N = 5) averaged probability of alert and sniff time ± 95% confidence interval to the different testing odors and to whole human scent during training session in between testing sessions.
Data points with different superscripts within a panel are statistically different from each other. A. shows dogs’ probability of alert to the different testing odors. B) Shows dogs sniff time (s) to the testing odors. Dashed line indicates the 3 s mark which was our alert criterion. Sniff time above 3 seconds indicated that dogs sampled the ports multiple times.