Making scents: dynamic olfactometry for threshold measurement

Abstract

Data on human odor thresholds show disparities huge enough to marginalize olfactory psychophysics and delegitimize importation of its data into other areas. Variation of orders of magnitude from study to study, much of it systematic, threatens meaningful comparisons with animal species, comparison between in vivo with in vitro studies, the search for molecular determinants of potency, and use of olfactory information for environmental or public health policy. On the premise that good experimental results will flow from use of good tools, this report describes a vapor delivery system and its peripherals that instantiate good tools. The vapor delivery device 8 (VDD8) provides flexibility in range of delivered concentrations, offers definable stability of delivery, accommodates solvent-free delivery below a part per trillion, gives a realistic interface with subjects, has accessible and replaceable components, and adapts to a variety of psychophysical methodologies. The device serves most often for measurement of absolute sensitivity, where its design encourages collection of thousands of judgments per day from subjects tested simultaneously. The results have shown humans to be more sensitive and less variable than has previous testing. The VDD8 can also serve for measurement of differential sensitivity, discrimination of quality, and perception of mixtures and masking. The exposition seeks to transmit general lessons while it proffers some specifics of design to reproduce features of the device in a new or existing system. The principles can apply to devices for animal testing.

Figure 1

A schematic shows essential parts of the VDD8. Generation of vapor begins with flow of inert nitrogen (feed stream) through a MFC to a heater that receives a cross-flow of liquid from a syringe. The vapor then goes into the 1.9-L vapor capacitor (larger cylinder). The vapor may then go through an Attenuator to dilute it one or 2 stages (up to 800 000:1) or may bypass the Attenuator. When the vapor enters the distribution manifold, it splits into 8 (or 4) lines, each to 1 cone of the 3 in a station. (The 8-path distribution manifold can become 2 four-path manifolds operated independently. The alternate vapor generator refers to a setup that duplicates the components outlined by the dashed line.) Just below where flow enters a cone, a fitting allows vapor sampling. The flow of vapor enters the bottom of a cone where it mixes with a background flow of air provided by a regenerative blower (oil-less ring compressor). All cones receive the same flow of air, typically 40 L/min, cleaned by activated carbon just before it enters a cone. A perforated disk in each cone creates turbulence to promote mixing. The mouth of the cone affords a third place to sample vapor concentration. The photo inset gives a sense of scale, with the 8-rotameter unit distribution manifold, the 4-rotameter unit Attenuator above it, and the 4-rotameter unit background odorizer. This figure appears in color in the online version of Chemical Senses.

Figure 2

Upper part: spreadsheet to set up the VDD for a given outcome, in this case for D-limonene at a maximum concentration of 100 ppb and 2-fold dilutions over 8 stations. The user enters the information in the left columns and the spreadsheet returns the information in the right columns. Under VDD settings, the information with asterisks represents that customarily used to generate data for a psychometric function. The entry of 1 for Attenuation provides a starting point that may need adjustment. Under calculated properties, the asterisked information (top cell) lies below the nominal minimum for the liquid feed rate from the syringe. In such a case, the operator can increase the entry for Attenuation. Ignoring that for the moment, the calculated properties pose no other problems. The “dynamic range” of 128:1 merely represents 7 successive halvings from the highest concentration. The calculation for Maximum Feed w/o Condensation shows that the feed rate of 0.05 μL/min lies very far from a rate that would cause “condensation.” Hence, the spreadsheet has returned the answer “No.” With the entry of 1 for Attenuation, the “concentration in vapor capacitor” and concentration to cones both equal the same value, in this case 2 ppm. Assuming accurate calibration of the 8 rotameters of the distribution manifold, the values in the table of concentration at cones, that is, 100 ppbv, 50 ppbv, etc., should hold as well. Lower part: the lower spreadsheet differs from the upper in small but essential ways. Because the upper indicated a liquid feed rate below the critical value for uniform delivery, the operator entered 20 into Attenuation. With the calculated liquid feed rate of 1.07 μL/min, approximately 4 times the critical value, the only other change in the lower sheet appears in concentration in vapor capacitor, where concentration has increased by 20-fold. The Attenuator, designed to dilute concentration, then comes into service. This figure appears in color in the online version of Chemical Senses.

Figure 3

Showing a version of an interactive spreadsheet for use with a VOC of very low threshold and slight solubility in water (ethyl n-butyrate) and a solvent of water instead of predilution with nitrogen. As do the spreadsheets in Figure 2, this has cells for user input. Rather than a cell for Attenuation, the sheet has a cell concentration of solute. In this case, where the solution loaded into the syringe contains 99.995% water, the water vapor concentration in the stream would limit the maximum feed rate of the syringe pump. At the Liquid Feed Rate calculated to deliver a maximum VOC concentration of 0.25 ppbv, the concentration of water vapor in the stream (735 ppm) lies at only 3.2% of its saturated vapor concentration. This figure appears in color in the online version of Chemical Senses.

Figure 4

Showing 3 examples of calibration of analytical instruments and validation of delivery for the VDD8. “Top row” shows calibration of response from a GC-ECD to liquid injections (0.5 μL) of chloropicrin in n-heptane and validation of delivery for 250-μL vapor samples from the cones. The average CV of the direct vapor samples equaled 10%. “Middle row” shows calibration of response from an HPLC to liquid injections (20 μL) of glutaraldehyde-bis-DNPH in acetonitrile and validation of delivery with injected liquid samples of the same reaction product (derivative) obtained after trapping glutaraldehyde onto treated filters and reacting the trapped material with DNPH and phosphoric acid. CV equaled 10%. “Bottom row” shows calibration of response from a GC-FID to liquid injections (0.5 μL) of ethyl n-butyrate in ethanol and validation of delivery from thermally desorbed vapor samples collected from the cones onto Tenax.

Figure 5

Showing how well 4 young subjects (3 males and a female) detected the tutti-frutti odor of ethyl butyrate in 3-alternative forced-choice testing, with concentrations down to 2 ppt. The subjects gave informed consent to participate in a protocol approved by an Institutional Review Board of the University. Each contributed 100 judgments per point over 3 days of testing. Threshold occurred at an average of 15 ppt, 4 orders of magnitude below that listed in the Handbook of Industrial Toxicology and Hazardous Materials (Cheremisanoff 1999). For details of protocol, such as timing, see appendix (Supplementary Material).