Rats Smell in Stereo

See allHide authors and affiliations

Science  03 Feb 2006:
Vol. 311, Issue 5761, pp. 666-670
DOI: 10.1126/science.1122096


It has been hypothesized that rats and other mammals can use stereo cues to localize odor sources, but there is limited behavioral evidence to support this hypothesis. We found that rats trained on an odor-localization task can localize odors accurately in one or two sniffs. Bilateral sampling was essential for accurate odor localization, with internasal intensity and timing differences as directional cues. If the stimulus arrived at the correct point of the respiration cycle, internasal timing differences as short as 50 milliseconds sufficed. Neuronal recordings show that bulbar neurons responded differentially to stimuli from the left and stimuli from the right.

Rats use olfactory cues to locate and identify objects in their environment (1, 2). Odor sources can be localized by one of two broad mechanisms (3): sequentially comparing odor concentrations at two different locations (4) or comparing simultaneous samples from two different locations of the body (57). The latter strategy requires separate sampling and parallel neuronal pathways that eventually converge for bilateral comparison. Rat nostrils are about 3 mm apart and at first sight appear to be too close to each other to support separate sampling. However, studies of respiratory air-flow patterns have shown that air flow is directed laterally to the left and right of the respective nostrils, which suggests that separate sampling of the olfactory environment may occur (8). Further, the two nasal passages are almost completely isolated from each other and supply two distinct sheets of olfactory sensory epithelia. The axonal projections from the sensory epithelium maintain this separation into the first olfactory region of the brain, the olfactory bulb. Higher olfactory cortical areas are sites of bilateral convergence. Consistent with this bilateral convergence, neurons in the anterior piriform cortex have been shown to respond differentially to ipsilateral, contralateral, or bilateral odor stimulation (9). Thus, the olfactory system of the rat appears to satisfy the two requirements of separate sampling and neuronal mechanisms for stereo odor localization. We therefore tested the hypothesis that rats localize odor sources using stereo cues.

Using standard behavioral conditioning procedures, we trained water-deprived rats to localize the source of an odor to the left or to the right. Each rat initiated a trial by poking its nose into a sniff port. After a random interval of 0 to 100 ms, an odor stimulus was delivered on either the left or the right. Rats received water rewards for licking at one of two water spouts: the left water spout for a left odor trial and the right water spout for a right odor trial. Rats performed 100 to 200 trials per session, and one session per day, for consecutive days except weekends. Figure 1A shows the four stages of a trial: nose poke, odor onset, nose withdrawal, and lick. The odor stimulus of intensity 1% of saturated vapor was produced using a custom-made air-dilution olfactometer. Air flow was maintained at 5 liters/min on both sides throughout a session. Each of the rats was trained on one of three odors: isoamyl acetate, 1-4 cineole, or phenethyl alcohol.

Fig. 1.

Accurate odor localization requires one to two sniffs. (A) Odor-localization task. The four stages of the odor-localization task—nose poke, odor onset, nose withdrawal, and lick—are shown here. The rat received a water reward if it licked the spout on the same side as the odor source. The sampling times and movement times were measured as the time from odor onset to nose withdrawal and from nose withdrawal to lick, respectively. (B) Acquisition of the task. Accuracy as a function of training days for three rats. Day 1 corresponds to the start of training on the actual task [stage 5 (14)]. (C) Distribution of sampling times over five sessions for one rat performing the task with isoamyl acetate as the odor. The sampling times for the left-side odor trials and right-side odor trials are shown separately. The empty bars indicate correct trials, and the filled bars indicate error trials. The median sampling times for the left and the right were 314.8 ms and 332.6 ms, respectively. (D) Distribution of movement times for the same rat. The median movement times for the left and the right were 316.8 ms and 280.9 ms, respectively. (E) Accuracies of the rats on different controls. These control sessions were randomly performed between days with normal sessions. All the days were significantly different from the “No nitrogen flow” control session (**, ANOVA, P < 0.0005). (F) Number of sniffs taken during left and right odor trials for two thermocouple-implanted rats (five sessions for each rat). Tubing delays account for ∼100 ms, or about 1 sniff.

We trained 14 Wistar rats (12 male and 2 female) to a performance criterion of 80% accuracy per session on the odor-localization task (Fig. 1B). Figure 1, C and D, shows the distribution of odor sampling times and movement times for one rat over five sessions. Median odor sampling times for individual training sessions were in the range of 280 ms to 400 ms (10).

To ensure that the rats used only odor cues to solve the task, we performed a number of controls to address other cues: (i) Absolute intensity differences. We tested the rats with the left and right odor channels interchanged. Their performance was not affected by this change (n = 5). (ii) Sound cues. We tested the rats with no flow through the odor bottles, although the sound cues due to valve switching remained. Performance of the rats dropped to chance levels (n = 5). (iii) Flow rate changes as somatosensory cues. Odor onset was associated with a 1% change in total air-flow rate. It has been hypothesized that air flow can be detected by the whiskers (11). We trimmed the anterior sets of whiskers and found that this did not affect the performance of the rats (n = 6). Further, differences of 1% in flow between the two sides were smaller than the error range of the flow meters and so are unlikely to be consistent cues. (iv) Air-flow transients. Air-flow transients and pressure pulses propagate at the speed of sound, whereas the odor itself propagates at the speed of the air flow in the tubes (∼7 m/s). We tested the rats (n = 6) with longer odor tubes connected to the behavior box and found a significant increase in sampling time that matched our estimates of air-flow rates (ANOVA, P < 0.05) (fig. S2). We also tested the rats on sessions where odor was already present on one of the sides before a nose poke, thus eliminating transients (n = 5). Rats performed above criterion on such sessions. Figure 1E shows the accuracy of the rats on the different control sessions.

We then implanted thermocouples in the nostrils of three of the rats to monitor respiration in terms of temperature changes in inhaled and exhaled air (fig. S3). Similar to earlier studies (12, 13), we found that rats typically sniffed at 7 to 8 Hz during odor sampling (fig. S3). Odorant direction discrimination was complete within two to three sniffs on most trials (Fig. 1F). We recorded odor onset as the time of application of a switching signal to the electronic valve controlling odor flow. Therefore, the above sampling times include the time taken for the odor to reach the rat. We estimate that the odor takes ∼80 to 100 ms to reach the rat after the switching signal (14). Taking this into account, our results suggest that rats localize odor sources accurately, with odor sampling times of ∼250 ms, within one to two sniffs.

To determine whether the rats were using stereo cues to localize the source of the odor, we disrupted bilateral odor sampling by stitching one of the nostrils shut (Fig. 2A). The measurements started after rats performed at criterion for several days. Performance was monitored in training sessions for 13 days. On days 5 and 6 the right nostril was blocked by stitching, and on days 10 and 11 the left nostril was blocked. Nostril blocking and unblocking were done in a 10-min surgical procedure under anaesthesia immediately after the training session on the preceding day. Training sessions immediately before, during, and immediately after the nostril-stitching procedure were always on consecutive days. Stitches reliably blocked air flow through the nostril on the first day of stitching (14). For some of the rats, stitches only partially blocked air flow on the second day.

Fig. 2.

Bilateral odor sampling is essential for accurate odor localization. (A) Experimental design. In all of these plots, the shaded region depicts the days on which one of the nostrils was stitched. The black line represents average values, and each symbol represents an individual rat. Nostril stitching was a 10-minute procedure done under anaesthesia. (B) Accuracy on the odor-localization task (filled symbols and, isoamyl acetate rats; open symbols + and X, 1-4 cineole rats). The first day of stitching in both cases was significantly different from the normal days (**, ANOVA, P < 0.0005), and the second day was significantly different from the normal days (*, ANOVA, P < 0.05). (C) Response bias on the odor-localization task. Symbols for individual rats are the same as in (B). Response bias was calculated as described in (21). A response bias of +2 indicates that the rat licked on the left for all the trials, and a response bias of –2 indicates that the rat licked on the right for all the trials. Response bias was not significantly different throughout as different rats adopt different strategies. (D) Accuracy on the odor-discrimination task. In this odor-discrimination task, one of two odors was delivered from either of the sides in the same apparatus. Rats were rewarded for licking on the left side for one of the odors and on the right side for the other odor, irrespective of the direction of the odor. Accuracies were not significantly different on any of the days. (E) Response bias on the odor-discrimination task. Response bias was not significantly different for any of the days. (F) Accuracy for two rats localizing phenethyl alcohol, a pure olfactory stimulant.

Figure 2B shows the accuracy of six rats that were used for the stitching experiments. Performance drops significantly below criterion when either of the nostrils is stitched shut and recovers immediately after removal of the stitches. The responses of most rats were biased to the unstitched side when one of the nostrils was stitched shut (Fig. 2C).

To exclude the possibility that poor performance could be a result of discomfort due to stitching, we trained two rats to perform a forced choice task where the identity, not the direction, of the odor was the cue. The rats were required to lick on the left water spout for one of the odors and lick on the right water spout for the other odor, irrespective of the side on which the odor was delivered. Performance on this task was not affected when one of the nostrils was stitched shut (Fig. 2D). In addition, the stitched rats did not exhibit any significant response bias (Fig. 2E).

The nasal cavity is innervated both by olfactory sensory neurons and by the ethmoid branch of the trigeminal nerve. The trigeminal nerve senses irritation and is activated by most, but not all, odorants at high concentrations. Studies on odor localization in humans have provided mixed results. Although von Békésy suggested that humans could use stereo cues (15), others suggested that only odorants that stimulated the trigeminal nerve could be localized effectively (16, 17). A recent study has provided evidence that, under controlled conditions, humans can localize pure olfactory stimulants (18). To determine whether rats could localize pure olfactory stimulants, we used phenethyl alcohol, an odorant that does not stimulate the rat trigeminal nerve (19), for two of the rats (Fig. 2F). Rats successfully localized phenethyl alcohol and required bilateral sampling for accurate localization.

The above results suggest that rats can use stereo cues to localize odor sources. By analogy with the auditory system (20), internostril intensity and timing differences are candidate stereo cues. In a modified task, odor was present on both sides, with the onset of odor on one side leading the onset of odor on the other side by a fixed interval (Fig. 3A). Rats were rewarded for choosing the side on which onset of odor occurred first. To avoid possible odor quality differences, we designed the olfactometer to use the same odor bottle for both sides. Therefore, the odor intensity dropped by half on the first side after the onset of odor on the other side (Fig. 3A). Thus, this task presented a brief intensity difference cue along with the timing difference cue. Alternatively, the task can be thought of as a unilateral stimulus followed by a bilateral masking stimulus after a fixed interval. In a given session, trials with intervals of 0, 25, 50, 100, and 150 ms were randomly interleaved. Performance on this task was above criterion for 100-ms and 150-ms intervals. We plotted the fraction of responses on the left side as a function of the interval between onsets of odor on both sides, negative intervals implying that onset of odor on the right side occurred first (Fig. 3B). These data were fitted to a psychometric function, a function that measures stimulus detectability as a function of small changes in stimulus quality. A threshold interval of 125 ms for performance above 80% was calculated from this function as the smallest interval that corresponded to 80% accuracy (Fig. 3B).

Fig. 3.

Internostril intensity and timing differences can be used as cues. (A) Experimental design. We used the same bottle to deliver odor on both the left and the right to avoid qualitative differences. This meantthattheintervaltask presented a brief intensity difference along with the timing difference. (B) Response to odor timing differences. Each symbol depicts the accuracies of one rat (n = 6) on the different intervals used in the interval task. The black line represents the psychometric function that was fit to the average of all the rats. (C) Respiration phase dependence of accuracy for each of the intervals. The plot shows average accuracy and standard deviation for the three thermocouple-implanted rats. Two of the rats were trained on isoamyl acetate, and one of the rats was trained on 1-4 cineole. At long intervals, responses are accurate at all phases. At short intervals, responses are phase dependent and binning differences occur [see (D)]. (D) Dependence of accuracy on respiration phase for the 50-ms interval trials for all of the three rats implanted with thermocouples. The first bin in (C) and (D) corresponds to the beginning of inhalation.

Because the time scales of neuronal circuits are typically faster than 125 ms, we wanted to see whether this threshold for accurate performance was a result of intermittent sampling due to respiration. We used the thermocouple-implanted rats to determine the phase of respiration at which the odor first arrived in the interval task. We divided the respiration into five phases and computed response accuracy for each phase and each interval (Fig. 3C). Accuracies for the longer intervals showed little dependence on respiration phase, whereas accuracies for the shorter intervals (25 and 50 ms) were dependent on respiration phase. Accuracies in some bins reached close to criterion levels even for these short intervals (Fig. 3D). Thus, taking into account the sampling constraints imposed by respiration, intervals as short as 50 ms can be used efficiently for localization of odor sources.

What is the neuronal substrate for odor localization? If our interpretation of separate air sampling from the two nostrils is correct, then the distinct sensory neuron projections to the olfactory bulb should result in unilateral olfactory responses. More complex responses may arise because there are feedback projections to the bulb, as well as reciprocal inhibitory projections from the contralateral bulb, routed through the anterior olfactory nucleus.

Using tetrodes, we recorded from single units in the main olfactory bulb of anaesthetized female rats. We presented odor from either the left or the right in a randomly interleaved sequence, using the same olfactometer used for the behavior. Figure 4A shows the spike rasters and peristimulus time histograms for the response of a neuron in the left olfactory bulb, sorted by odor direction. Response significance was calculated by comparing firing rate during the odor period to the immediately preceding air period of the same duration. Twenty-one of 41 neurons showed a significant response to isoamyl acetate (one-way ANOVA, P < 0.05). Fifty-two percent of the responsive neurons (11/21) were unilaterally responsive, that is, they showed significant firing-rate changes to odor from one side only. Eighty percent (8/10) of the remaining, bilaterally responsive neurons were direction selective and exhibited significantly different spike rates during the left and right odor periods (one-way ANOVA, P < 0.05) (fig. S7). Overall, our recordings suggest that more than 90% of the responsive neurons in the olfactory bulb respond differentially to stimuli presented on the left or the right (Fig. 4, B and C). This could possibly provide a substrate for the nostril-specific activation seen in the olfactory cortex in a recent study on humans performing an odor-localization task (18).

Fig. 4.

Direction-selective responses of olfactory bulb neurons in anaesthetized, freely breathing female rats. (A) Raster and peristimulus time histograms for the response of a single unit in the left olfactory bulb to odor (isoamyl acetate) presented from the left and right, respectively. The black bar denotes the duration of odor presentation (4 s). Left- and right-side trials were randomly interleaved. (B) Classification of bulbar neurons: unilateral (11/41), bilateral direction selective (8/41), bilateral direction unselective (2/41), and unresponsive (20/41). Unilateral neurons showed significant firing-rate changes during odor (as compared with a preceding air period of the same duration, ANOVA, P < 0.05) only for odor from one of the sides. Bilateral neurons showed significant firing-rate changes in response to odor from either of the sides. Further, bilateral direction-selective neurons showed significant differences between the two odor periods (left and right, ANOVA, P < 0.05) (C) Distribution of direction selectivity. Strength of response (defined as firing rate during odor from specified direction/firing rate during preceding air period) on the right is plotted against strength of the response on the left for the responsive neurons. A log-log scale has been used. The 45° dashed line represents a lack of odor direction selectivity.

Our behavioral and recording results provide constraints on the neuronal mechanisms mediating stereo olfaction: (i) We show that odor localization can use stereo cues, implying bilaterally distinct neuronal pathways. This is supported by our recordings. (ii) Direction discrimination can be made within one sniff (125 ms), implying that the brain is performing a simultaneous rather than sequential comparison. (iii) The underlying neuronal circuits appear to be able to perform the discrimination more rapidly (50 ms) but are limited by the physiological sampling rate imposed by sniffing.

von Békésy reported that humans can localize odor sources using stereo cues (15) but estimated a timing selectivity of 100 μs, which is three orders of magnitude smaller than what we see in rats. Based on our estimates of ∼100-ms timing discrimination and an effective spacing of ∼1 cm between the air-sampling regions for each rat nostril, odor plumes can be localized if they traverse the nostrils at less than 10 cm/sec. Relatively laminar air flow may lead to more sustained gradients that the rat may be able to track, through stereo sampling, to an upstream odor source. Stereo localization in a single sniff is at least twice as fast as sequential sampling, which may be important both for foraging and for predator avoidance. We suggest that, for a rat, each sniff is a perceptually complete snapshot of the olfactory world, including both odor identity (12) and stereo-based location.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

References Movie S1

References and Notes

View Abstract

Navigate This Article