- Email: [email protected]
Behavioral Responses to Aliphatic Aldehydes Can Be Predicted From Known Electrophysiological Responses of Mitral Cells in the Olfactory Bulb
Physiology & Behavior, Vol. 66, No. 3, pp. 497–502, 1999 © 1999 Elsevier Science Inc. Printed in the USA. All rights reserved 0031-9384/99/$–see front matter
PII S0031-9384(98)00324-2
Behavioral Responses to Aliphatic Aldehydes Can Be Predicted From Known Electrophysiological Responses of Mitral Cells in the Olfactory Bulb CHRISTIANE LINSTER1 AND MICHAEL E. HASSELMO Department of Psychology, Harvard University, 33 Kirkland Street, Cambridge, MA 02138 Received 28 July 1998; Accepted 16 November 1998 LINSTER, C. AND M. E. HASSELMO. Behavioral responses to aliphatic aldehydes can be predicted from known electrophysiological responses of mitral cells in the olfactory bulb. PHYSIOL BEHAV 66(3) 497–502, 1999.—For a better understanding of the encoding of odor quality in the olfactory system, it is critical to determine how electrophysiological responses to odorants are reflected in the behavioral responses to these odorants. In this article, we use a simple behavioral paradigm to show that the behavioral responses to similar odorants can be predicted from the electrophysiological responses of neurons in the olfactory bulb. Carbon chain length in aliphatic aldehydes has been used as a model for graded similarity among odorants. Recent electrophysiological experiments have shown that mitral cells in the rabbit olfactory bulb respond with similar response patterns to aliphatic aldehydes of similar chain length. On average, mitral cells responded with increased spiking activity to stimulation with two to three different aldehydes of neighboring chain length. We here show that the perception of these odorants can be predicted from the electrophysiological responses: rats that are conditioned to a given aldehyde generalize to aldehydes with one to two carbon differences in chain length from the conditioned aldehyde. When asked to discriminate between aldehydes of different chain lengths, rats learned to discriminate between any two odorants, but the rate of acquisition depended on the degree of similarity between the two odorants. © 1999 Elsevier Science Inc. Aliphatic aldehydes
Electrophysiological responses
Mitral cells
Olfactory bulb
fatty acids, and aliphatic alcohols of various chain lengths as models for graded similarity among odorants (2–4,9,11,13). In particular, these studies have shown that in the olfactory bulb, within a given group of compounds, mitral cells tend to respond preferentially to odorants with similar carbon chain lengths. As a result, the overlap in mitral cell responses between two “similar” aldehydes (i.e., with a similar number of carbons in the chain) would be larger than it would be for two more dissimilar aldehydes (3,13). We predicted that the large overlap in neural responses to “similar” odorants, as seen in the olfactory bulb, should manifest as a greater perceptual similarity of these odorants. Our behavioral protocol was adapted from Bunsey and Eichenbaum (1), and has been previously employed in our laboratory to study generalization between binary odorant
ONE of the fundamental questions in olfaction concerns the perceptual similarity between odorants. Electrophysiological studies in several species (3,4,9,10–13) suggest that neural response profiles (in the olfactory mucosa as well as in the olfactory bulb) correlate with features of the odorant molecules. However, the relationship between the neural responses in the olfactory system and the animal’s perception, as measured by a behavioral response, has not been directly investigated. We use a simple olfactory generalization paradigm to test whether behavioral responses to a group of structurally related odorants, unbranched aliphatic aldehydes, can be predicted from the known electrophysiological responses of olfactory bulb neurons to these same odorants. Recently, several electrophysiological studies in rabbit and mouse have used unbranched aliphatic aldehydes, aromatic
1To whom requests for reprints should be addressed at Department of Psychology, Boston University, 34 Cummington Street, Boston, MA 02215. E-mail: [email protected].
497
498
LINSTER AND HASSELMO
mixtures and their components (7,8). Rats were trained to retrieve a reward from a cup filled with bedding. We trained rats on a series of aliphatic aldehydes (Table 1), and tested pairwise generalization between these different odorants. Generalization was quantified as the amount of time spent digging in response to a test odor relative to the response to the conditioned odor. The degree to which rats “confuse” the test odorant with the conditioned odorant is a measure for the generalization between the two odorants, i.e., a measure for similarities in perception. To control for nonspecific responses to odorants, we also tested rats’ responses to an unrelated control odorant, n-amyl acetate. METHODS
Subjects Nine male Sprague–Dawley rats, weighing 400–450 g at the beginning of training, served as subjects for both the generalization and the discrimination experiment. Water was continuously available, but subjects were maintained on a food deprivation schedule designed to keep them at approximately 90% of their ad lib body weight during the experimental sessions. They were maintained on a 12 L:12 D cycle in an environmentally controlled room. All behavioral training was conducted near the end of the light cycle (1700 h). Apparatus All behavioral training took place in a transparent Plexiglas chamber divided into two subchambers by a sliding, opaque Plexiglas board (Fig. 1A). Odors were placed at the bottom of a glass cup (10-cm diameter, 5 cm high). The tip of a fresh Q-tip was covered with a fine plastic mesh that was taped to the bottom of the cup. At the beginning of each training set, the Q-tip was saturated with a 0.1-mL drop of diluted odor (5% vol/vol in mineral oil). The cup was then filled with bedding. The reward, bits of sweetened cereal (Kellogg’s Froot Loops), were buried in the bedding. After each trial, the bedding in the glass dish was replaced. Procedure Rats were shaped first to retrieve the reward by digging through the bedding. At the beginning of each trial, the rat was placed in subchamber 1, and the separation between the subchambers was closed. When the separation door was opened, the rat entered subchamber 2, and was allowed to dig in both cups until it retrieved the reward. Placement of the two cups was random and utilized the whole space of subchamber 2 in such a way that spatial arrangement of the cups had no predictive value as to the placement of the reward. The rats were trained to return to subchamber A after retrieving the reward or to signal the end of the trial. In all cases the
trial was ended by the experimenter after a maximum of 2 min. The odors were coded before the experiment in such a way that the experimenter did not know the identity of the odorants. Generalization Task Each training set consisted of eight initial training trials (rewarded) followed by three test trials (unrewarded) (Fig. 1B). Rats were first conditioned to a given aldehyde during eight training trials in which they had a choice between a scented cup containing the reward and an unscented cup. After the eight conditioning trials, three test trials (in randomized order) were performed, separated by one or two conditioning trials (Fig. 1C). During the test trials the rats had a choice between a scented cup containing the test odor but no reward, and an unscented cup. During these unrewarded test trials, the digging times in response to the conditioned aldehyde (conditioned odor), a second aldehyde of different chain length (test odor), and the control odor (n-amyl-acetate) were recorded. Each rat was tested with all possible pairs of aldehydes (the eight aldehydes used are shown in Table 1), one as conditioned odor, and one as test odor. On each day, the combination of conditioned odor and test odor was assigned randomly to each rat. Thus, each rat was tested with 30 possible combinations in a pseudorandomized order. The experimenter did not know the identity of the test odors during these experiments. The control odor, n-amyl acetate, served as a control for nonspecific generalization. Discrimination Task In this experiment, rats were presented with two cups; each contained a different odorant (A and B), but only one of the cups contained the reward. Each training set was composed of 10 consecutive trials with the same two odorants. During each trial, we recorded in which cup digging was first observed , but subjects were left to dig in either cup until the reward was retrieved. On each day and for each subject, a different combination of odorants was used, and contingency was assigned randomly. We tested the discrimination between [4]CHO and all other aliphatic aldehydes shown in Table 1 (five combinations). Statistical Analyses All statistical analyses were performed using JMP statistical software on mean time (in seconds) of digging activity, which are the values shown in most figures (6standard error). Three-way ANOVA was used to analyze the time each subject spent digging in the scented dish on each test trial. Subsequently, t-tests were performed for individual comparisons
TABLE 1 LIST OF ODORANTS USED
ALIPHATIC ALDEHYDES
499
FIG. 1. Experimental setup and protocol. (A) Behavioral training took place in a transparent Plexiglas chamber divided into two subchambers (1 and 2) by a sliding, opaque Plexiglas board. At the beginning of each trial the rat was placed in subchamber 1 while the separating door was closed. Two cups containing bedding, one of which contained both the odor and the reward, the other containing no odor and no reward, were then placed in subchamber 2. Placement of the two cups was randomly varied using the whole space of subchamber 2 in such a way that spatial arrangement of the cups had no predictive value as to the placement of the reward. The separation door was opened, and the rat entered chamber 2. After retrieval of the reward, the rats were returned to subchamber 1 and the door was closed. Opening of the door signaled the beginning of the next trial. (B) Protocol (generalization task): each training set consisted of eight initial training trials with the conditioning odor (rewarded) followed by three test trials (unrewarded) in randomized order. The test trials consisted of the conditioning odor, the test odor, and the control odor. The three test trials were separated by a small randomized number (0–3) of rewarded training trials. Duration of digging in the scented cup was measured with a stopwatch during test trials.
between the responses to the conditioning odor, the test odor, and the control odor. Thus, the ANOVA returned five F-ratios. The first three indicated differences among main treatment effects: 1) differences among rats in odor preferences (FRAT); 2) differences in the time spent digging across the three types of odors—conditioning odor, test odor, and control odor (FODOR); and 3) the effect of the difference in carbon chain lengths between the conditioning odor and the test odor (FDIFF). The two remaining F-ratios indicated interaction effects: 1) the interaction between subjects and test odorants (FRAT-ODOR) indicates how subjects differ in the relative response to the three test odors; and 2) the interaction between the test odors and the difference in carbon chain length (FODOR-DIFF) indicates if the relative response pattern to the three odorants depends on the difference in chain length between the conditioned odor and the test odor. RESULTS
Generalization Task Rats rapidly acquired each training set. After conditioning to a given odorant, rats responded strongly to that odorant in
the forced-choice test procedure in which they were presented with a choice between a dish scented with the conditioned odor versus an unscented dish (Fig. 2A, gray bars). On average, the longest digging times were observed in response to the conditioned odorant. When confronted with a choice between an aliphatic aldehyde different from the conditioned odor and an unscented dish (Fig. 2A, black bars), average digging times were substantially lower than in response to the trained odor. The responses were always highly dependent on the similarity between the conditioned odor and the test odor. On average, rats generalized only to aldehydes differing by one or two carbons from the conditioned odor; in all other cases, the response to the test odor was similar to the response to the control odor (n-amyl-acetate; Fig. 2A, white bars). ANOVA indicated that individual rats differed in their overall response levels to odors, FRAT 5 9.35, p , 0.0001, but there was no significant interaction effect between subjects and their relative response to the three odorants (FRAT-ODOR), indicating that the difference between subjects was mostly due to overall differences in response levels rather than to differences in their response patterns. However, the interaction between differences in chain length and the relative re-
500
LINSTER AND HASSELMO
sponse to conditioning versus test odor was highly significant, FODOR-DIFF 5 3.86, p , 0.0005, showing that the pattern of response to conditioning and test odorant was dependent on the relationships between the aldehydes used for conditioning and testing in a given trial. The average response levels to test odors differing by a single or two carbon chains from the conditioned odor were significantly higher (by two-way analysis of variance) than the response levels to the control odor (difference of a single carbon: F 5 69.25, p , 0.0001; difference of two carbons: F 5 7.35, p , 0.0075); for all other odor combinations, the responses to the test odor and to the control odor were not significantly different. Discrimination Task The results from these generalization experiments suggest that rats perceive groups of aldehydes similarly, as expressed by prolonged digging in a dish scented with an odorant, which is structurally similar to the conditioned odorant. This raises a question regarding the discriminability of these similar odorants. We, therefore, tested, in a second experiment, whether the same rats could learn to discriminate between these odorants. In this experiment, rats had to discriminate between two scented cups containing aliphatic aldehydes, over 10 successive trials. Specifically, we tested the rats’ ability to discriminate between [4]CHO and all other aliphatic aldehydes used in the previous experiment (Table 1). Figure 3A shows the average acquisition curves for all rats (n 5 9). Discrimination tasks involving very dissimilar aldehydes were solved rapidly
by all subjects (100% correct after three to four trials), while when more similar odors were presented, the rats did not reach maximal performance until a much higher number of trials (80% correct at the eighth trial). We have shown that odor pairs differing by only one carbon are more difficult to discriminate from one another than odor pairs differing by two, three, or four carbons. The question remains whether this degree of difficulty correlates with the differences in the number of carbons per se, regardless of the absolute number of carbons in either molecule. We addressed this question by correlating all possible pairs of task acquisition curves (from Fig. 3A) with one another. The resulting correlation coefficients were then grouped by the difference in the difficulties of the two task acquisition curves that were correlated; for example, the two tasks “[4]CHO versus [3]CHO” and “[4]CHO versus [5]CHO” have the same difficulty because the difference in chain length between the two odorants is one in each case. The two tasks “[4]CHO versus [3]CHO” and “[4]CHO versus [8]CHO,” however, are of very different difficulty, because in one task the difference in carbon chain length between the two odorants is one, and in the other task it is four. The results (Fig. 3B) indicate that pairs of task acquisition curves derived from tasks of similar difficulty are more correlated with one another than are pairs of curves derived from tasks of dissimilar difficulty. For example, the acquisition curve “[3]CHO versus [4]CHO” was highly correlated (0.8) to the curve “[4]CHO versus [5]CHO” and the two curves were not significantly different (p . 0.2, paired t-test), while the correlation between the
FIG. 2. Generalization between aliphatic aldehydes of varying chain lengths. (A) The graph shows the average digging times [all rats (n 5 9) and all odor combinations (n 5 30)] in response to the conditioned odor (A), the test odor (B), and the control odor (X) as a function of the difference in chain lengths between the conditioned odor and the test odor. (B)The graph shows the average response levels [all rats (n 5 9)] to all test odors after conditioning with all six training odors ([3]CHO–[8]CHO). We compared the average response level to each test odor to the response to the control odor (two-way analysis of variance, *indicates p , 0.01).
ALIPHATIC ALDEHYDES
FIG. 3. Discrimination between aliphatic aldehydes of different chain length. (A) The graph shows the average acquisition curves for discrimination between [4]CHO and all other odorants used. (B) Correlation coefficients for the two by two comparisons of the acquisition curves shown in A. All the correlation coefficients are given. C(4 versus 3, 4 versus 5) is the correlation coefficient calculated for the task acquisition curves [4]CHO versus [4]CHO and [4]CHO versus [5]CHO. The correlation coefficients are grouped by the difference in difficulty of the two task acquisition curves. For example, [4] versus [6] and [4] versus [7] are relatively similar tasks (in one, the difference in chain length between the two odorants is 2; in the other, the difference in chain length is 3); whereas [4] versus [5] and [4] versus [8] are dissimilar tasks (in one, the difference in chain length between the two odorants is 1, and in the other it is 4). *Indicates when the two compared curves are significantly different from each other (p , 0.05, paired t-test).
501
502
LINSTER AND HASSELMO
curve “[4]CHO versus [5]CHO” and the curve “[4]CHO versus [7]CHO” was lower (0.5) and the two curves were significantly different (p , 0.01, paired t-test). Similarly, in the case of two “difficult tasks,” the acquisition curve “[6]CHO versus [8]CHO” was highly correlated to the curve “[7]CHO versus [8]CHO” and the curves were not statistically different (p . 0.2, paired t-test). All discrimination tasks could be learned by the rats given enough trials (after 10 trials, performance in all tasks increases to 100% correct responses). Thus, even “similar” odorants can be discriminated given appropriate motivation and acquisition time. The acquisition of the discrimination task could potentially have been influenced by the previous exposure to the aldehydes during the generalization task, as the same rats were used in both tasks. However, all the acquisition curves would have been affected similarly and the preexposure would be unlikely to influence the comparative results presented here. DISCUSSION
The results from the behavioral experiments presented herein suggest that the perceptual similarity of odorants can be predicted from the similarities in the neural responses evoked by these odorants in the olfactory bulb. Mitral cells in the olfactory bulb that are activated by aldehydes (among other odorants) tend to respond with increased firing rates to a range of aliphatic aldehydes differing by one or two carbons from the “best” odor (3,10,13). Similarly, in our experiments, rats generalized to aldehydes differing by one or two carbons from the conditioned odor. Thus, the large overlap in neural responses to these odorants as seen in the olfactory bulb is
predictive of a similar overlap in perception. The response to the conditioned odor is always at least threefold that of the test odors, suggesting that the overlap in representation leads to associative spread. This associative spread is behaviorally expressed as “confusion” between odorants. The fact that all rats learned to discriminate between similar aldehydes very rapidly (over a maximum of 10 trials) in the second experiment shows that, despite the generalization, the representations of these odorants are sufficiently different from each other to allow discrimination. Olfactory sensory neurons, presynaptic to mitral cells, respond to a larger number of compounds within a group of similar odorants than do individual mitral cells in the olfactory bulb (11). The relatively narrow response spectra of individual mitral cells to similar odorants within a group of related odorants is at least partly shaped by the inhibitory circuits in the olfactory bulb: when inhibition in the olfactory bulb is suppressed, mitral cells respond with increased firing rates to a larger number of compounds (13). We propose that generalization between similar components could arise from the overlap in responsiveness of mitral cells to these odorants, corresponding to a spread of association. In contrast, given appropriate motivation, learned discrimination between similar odorants necessitates the enhancement of the differences between the two patterns. Such an enhancement can arise from modulation of inhibition in the olfactory bulb during the learning task (5,6), or could arise from a self-organizing process in olfactory cortical areas. ACKNOWLEDGEMENTS
The authors thank Brian Smith for fruitful discussions and Thomas Cleland for helpful suggestions on the manuscript.
REFERENCES 1. Bunsey, M.; Eichenbaum, H.: Conservation of hippocampal memory function in rats and humans. Nature 379:255–257; 1996. 2. Cometto-Muniz, J. E.; Cain, W. S.; Abraham, M. H.: Nasal pungency and odor of homologous aldehydes and carboxylic acids. Exp. Brain Res. 118:180–188; 1998. 3. Imamura, K.; Mataga, N.; Mori, K.: Coding of odor molecules by mitral/tufted cells in rabbit olfactory bulb. I. Aliphatic compounds. J. Neurophysiol. 68:1986–2002; 1992. 4. Katoh, K.; Koshimoto, H.; Tani, A.; Mori, K.: Coding of odor molecules by mitral/tufted cells in rabbit olfactory bulb. II. Aromatic compounds. J. Neurophysiol. 70:2161–2175; 1993. 5. Linster, C.; Gervais, R.: Investigation of the role of interneurons and their modulation by centrifugal fibers in a neural model of the olfactory bulb. J. Comp. Neurosci. 3:225–246; 1996. 6. Linster, C.; Hasselmo, M.: Modulation of inhibition in an olfactory bulb model reduces overlap in the neural representation of olfactory stimuli. Behav. Brain Res. 84:117–127; 1997. 7. Linster, C.; Smith, B. H.: Generalization between binary odor mixtures and their components in the rat. Physiology and Behavior (in press).
8. Linster, C.; Smith, B. E.; Hasselmo, M. E.: Rats generalize between binary odor mixtures and their components. Am Chem Soc Meeting, Sarasota, FL; 1998. 9. Mori, K.; Mataga, N.; Imamura, K.: Differential specificities of single mitral cells in rabbit olfactory bulb for a homologous series of fatty acid odor molecules. J. Neurophysiol. 67:786–789; 1992. 10. Motokizawa, F.: Odor representation and discrimination in mitral/tufted cells of the rat olfactory bulb. Exp. Brain Res. 112:24–34; 1996. 11. Sato, T.; Hirono, J.; Tonoike, M.; Takebayashi, M.: Tuning specificities to aliphatic odorants in mouse olfactory receptor neurons and their local distribution. J. Neurophysiol. 72:2980–2989; 1994. 12. Scott, J. W.; Davis, L. M.; Shannon, D.; Kalan, C.: Relation of chemical structure to spatial distribution of sensory responses in the rat olfactory epithelium. J. Neurophysiol. 75:2036–2048; 1996. 13. Yokoi, M.; Mori, K.; Shigetada, N.: Refinement of odor molecule tuning by dendrodendritic synaptic inhibition in the olfactory bulb. Proc. Natl. Acad. Sci. USA 92:3371–3375; 1995.
PII S0031-9384(98)00324-2
Behavioral Responses to Aliphatic Aldehydes Can Be Predicted From Known Electrophysiological Responses of Mitral Cells in the Olfactory Bulb CHRISTIANE LINSTER1 AND MICHAEL E. HASSELMO Department of Psychology, Harvard University, 33 Kirkland Street, Cambridge, MA 02138 Received 28 July 1998; Accepted 16 November 1998 LINSTER, C. AND M. E. HASSELMO. Behavioral responses to aliphatic aldehydes can be predicted from known electrophysiological responses of mitral cells in the olfactory bulb. PHYSIOL BEHAV 66(3) 497–502, 1999.—For a better understanding of the encoding of odor quality in the olfactory system, it is critical to determine how electrophysiological responses to odorants are reflected in the behavioral responses to these odorants. In this article, we use a simple behavioral paradigm to show that the behavioral responses to similar odorants can be predicted from the electrophysiological responses of neurons in the olfactory bulb. Carbon chain length in aliphatic aldehydes has been used as a model for graded similarity among odorants. Recent electrophysiological experiments have shown that mitral cells in the rabbit olfactory bulb respond with similar response patterns to aliphatic aldehydes of similar chain length. On average, mitral cells responded with increased spiking activity to stimulation with two to three different aldehydes of neighboring chain length. We here show that the perception of these odorants can be predicted from the electrophysiological responses: rats that are conditioned to a given aldehyde generalize to aldehydes with one to two carbon differences in chain length from the conditioned aldehyde. When asked to discriminate between aldehydes of different chain lengths, rats learned to discriminate between any two odorants, but the rate of acquisition depended on the degree of similarity between the two odorants. © 1999 Elsevier Science Inc. Aliphatic aldehydes
Electrophysiological responses
Mitral cells
Olfactory bulb
fatty acids, and aliphatic alcohols of various chain lengths as models for graded similarity among odorants (2–4,9,11,13). In particular, these studies have shown that in the olfactory bulb, within a given group of compounds, mitral cells tend to respond preferentially to odorants with similar carbon chain lengths. As a result, the overlap in mitral cell responses between two “similar” aldehydes (i.e., with a similar number of carbons in the chain) would be larger than it would be for two more dissimilar aldehydes (3,13). We predicted that the large overlap in neural responses to “similar” odorants, as seen in the olfactory bulb, should manifest as a greater perceptual similarity of these odorants. Our behavioral protocol was adapted from Bunsey and Eichenbaum (1), and has been previously employed in our laboratory to study generalization between binary odorant
ONE of the fundamental questions in olfaction concerns the perceptual similarity between odorants. Electrophysiological studies in several species (3,4,9,10–13) suggest that neural response profiles (in the olfactory mucosa as well as in the olfactory bulb) correlate with features of the odorant molecules. However, the relationship between the neural responses in the olfactory system and the animal’s perception, as measured by a behavioral response, has not been directly investigated. We use a simple olfactory generalization paradigm to test whether behavioral responses to a group of structurally related odorants, unbranched aliphatic aldehydes, can be predicted from the known electrophysiological responses of olfactory bulb neurons to these same odorants. Recently, several electrophysiological studies in rabbit and mouse have used unbranched aliphatic aldehydes, aromatic
1To whom requests for reprints should be addressed at Department of Psychology, Boston University, 34 Cummington Street, Boston, MA 02215. E-mail: [email protected].
497
498
LINSTER AND HASSELMO
mixtures and their components (7,8). Rats were trained to retrieve a reward from a cup filled with bedding. We trained rats on a series of aliphatic aldehydes (Table 1), and tested pairwise generalization between these different odorants. Generalization was quantified as the amount of time spent digging in response to a test odor relative to the response to the conditioned odor. The degree to which rats “confuse” the test odorant with the conditioned odorant is a measure for the generalization between the two odorants, i.e., a measure for similarities in perception. To control for nonspecific responses to odorants, we also tested rats’ responses to an unrelated control odorant, n-amyl acetate. METHODS
Subjects Nine male Sprague–Dawley rats, weighing 400–450 g at the beginning of training, served as subjects for both the generalization and the discrimination experiment. Water was continuously available, but subjects were maintained on a food deprivation schedule designed to keep them at approximately 90% of their ad lib body weight during the experimental sessions. They were maintained on a 12 L:12 D cycle in an environmentally controlled room. All behavioral training was conducted near the end of the light cycle (1700 h). Apparatus All behavioral training took place in a transparent Plexiglas chamber divided into two subchambers by a sliding, opaque Plexiglas board (Fig. 1A). Odors were placed at the bottom of a glass cup (10-cm diameter, 5 cm high). The tip of a fresh Q-tip was covered with a fine plastic mesh that was taped to the bottom of the cup. At the beginning of each training set, the Q-tip was saturated with a 0.1-mL drop of diluted odor (5% vol/vol in mineral oil). The cup was then filled with bedding. The reward, bits of sweetened cereal (Kellogg’s Froot Loops), were buried in the bedding. After each trial, the bedding in the glass dish was replaced. Procedure Rats were shaped first to retrieve the reward by digging through the bedding. At the beginning of each trial, the rat was placed in subchamber 1, and the separation between the subchambers was closed. When the separation door was opened, the rat entered subchamber 2, and was allowed to dig in both cups until it retrieved the reward. Placement of the two cups was random and utilized the whole space of subchamber 2 in such a way that spatial arrangement of the cups had no predictive value as to the placement of the reward. The rats were trained to return to subchamber A after retrieving the reward or to signal the end of the trial. In all cases the
trial was ended by the experimenter after a maximum of 2 min. The odors were coded before the experiment in such a way that the experimenter did not know the identity of the odorants. Generalization Task Each training set consisted of eight initial training trials (rewarded) followed by three test trials (unrewarded) (Fig. 1B). Rats were first conditioned to a given aldehyde during eight training trials in which they had a choice between a scented cup containing the reward and an unscented cup. After the eight conditioning trials, three test trials (in randomized order) were performed, separated by one or two conditioning trials (Fig. 1C). During the test trials the rats had a choice between a scented cup containing the test odor but no reward, and an unscented cup. During these unrewarded test trials, the digging times in response to the conditioned aldehyde (conditioned odor), a second aldehyde of different chain length (test odor), and the control odor (n-amyl-acetate) were recorded. Each rat was tested with all possible pairs of aldehydes (the eight aldehydes used are shown in Table 1), one as conditioned odor, and one as test odor. On each day, the combination of conditioned odor and test odor was assigned randomly to each rat. Thus, each rat was tested with 30 possible combinations in a pseudorandomized order. The experimenter did not know the identity of the test odors during these experiments. The control odor, n-amyl acetate, served as a control for nonspecific generalization. Discrimination Task In this experiment, rats were presented with two cups; each contained a different odorant (A and B), but only one of the cups contained the reward. Each training set was composed of 10 consecutive trials with the same two odorants. During each trial, we recorded in which cup digging was first observed , but subjects were left to dig in either cup until the reward was retrieved. On each day and for each subject, a different combination of odorants was used, and contingency was assigned randomly. We tested the discrimination between [4]CHO and all other aliphatic aldehydes shown in Table 1 (five combinations). Statistical Analyses All statistical analyses were performed using JMP statistical software on mean time (in seconds) of digging activity, which are the values shown in most figures (6standard error). Three-way ANOVA was used to analyze the time each subject spent digging in the scented dish on each test trial. Subsequently, t-tests were performed for individual comparisons
TABLE 1 LIST OF ODORANTS USED
ALIPHATIC ALDEHYDES
499
FIG. 1. Experimental setup and protocol. (A) Behavioral training took place in a transparent Plexiglas chamber divided into two subchambers (1 and 2) by a sliding, opaque Plexiglas board. At the beginning of each trial the rat was placed in subchamber 1 while the separating door was closed. Two cups containing bedding, one of which contained both the odor and the reward, the other containing no odor and no reward, were then placed in subchamber 2. Placement of the two cups was randomly varied using the whole space of subchamber 2 in such a way that spatial arrangement of the cups had no predictive value as to the placement of the reward. The separation door was opened, and the rat entered chamber 2. After retrieval of the reward, the rats were returned to subchamber 1 and the door was closed. Opening of the door signaled the beginning of the next trial. (B) Protocol (generalization task): each training set consisted of eight initial training trials with the conditioning odor (rewarded) followed by three test trials (unrewarded) in randomized order. The test trials consisted of the conditioning odor, the test odor, and the control odor. The three test trials were separated by a small randomized number (0–3) of rewarded training trials. Duration of digging in the scented cup was measured with a stopwatch during test trials.
between the responses to the conditioning odor, the test odor, and the control odor. Thus, the ANOVA returned five F-ratios. The first three indicated differences among main treatment effects: 1) differences among rats in odor preferences (FRAT); 2) differences in the time spent digging across the three types of odors—conditioning odor, test odor, and control odor (FODOR); and 3) the effect of the difference in carbon chain lengths between the conditioning odor and the test odor (FDIFF). The two remaining F-ratios indicated interaction effects: 1) the interaction between subjects and test odorants (FRAT-ODOR) indicates how subjects differ in the relative response to the three test odors; and 2) the interaction between the test odors and the difference in carbon chain length (FODOR-DIFF) indicates if the relative response pattern to the three odorants depends on the difference in chain length between the conditioned odor and the test odor. RESULTS
Generalization Task Rats rapidly acquired each training set. After conditioning to a given odorant, rats responded strongly to that odorant in
the forced-choice test procedure in which they were presented with a choice between a dish scented with the conditioned odor versus an unscented dish (Fig. 2A, gray bars). On average, the longest digging times were observed in response to the conditioned odorant. When confronted with a choice between an aliphatic aldehyde different from the conditioned odor and an unscented dish (Fig. 2A, black bars), average digging times were substantially lower than in response to the trained odor. The responses were always highly dependent on the similarity between the conditioned odor and the test odor. On average, rats generalized only to aldehydes differing by one or two carbons from the conditioned odor; in all other cases, the response to the test odor was similar to the response to the control odor (n-amyl-acetate; Fig. 2A, white bars). ANOVA indicated that individual rats differed in their overall response levels to odors, FRAT 5 9.35, p , 0.0001, but there was no significant interaction effect between subjects and their relative response to the three odorants (FRAT-ODOR), indicating that the difference between subjects was mostly due to overall differences in response levels rather than to differences in their response patterns. However, the interaction between differences in chain length and the relative re-
500
LINSTER AND HASSELMO
sponse to conditioning versus test odor was highly significant, FODOR-DIFF 5 3.86, p , 0.0005, showing that the pattern of response to conditioning and test odorant was dependent on the relationships between the aldehydes used for conditioning and testing in a given trial. The average response levels to test odors differing by a single or two carbon chains from the conditioned odor were significantly higher (by two-way analysis of variance) than the response levels to the control odor (difference of a single carbon: F 5 69.25, p , 0.0001; difference of two carbons: F 5 7.35, p , 0.0075); for all other odor combinations, the responses to the test odor and to the control odor were not significantly different. Discrimination Task The results from these generalization experiments suggest that rats perceive groups of aldehydes similarly, as expressed by prolonged digging in a dish scented with an odorant, which is structurally similar to the conditioned odorant. This raises a question regarding the discriminability of these similar odorants. We, therefore, tested, in a second experiment, whether the same rats could learn to discriminate between these odorants. In this experiment, rats had to discriminate between two scented cups containing aliphatic aldehydes, over 10 successive trials. Specifically, we tested the rats’ ability to discriminate between [4]CHO and all other aliphatic aldehydes used in the previous experiment (Table 1). Figure 3A shows the average acquisition curves for all rats (n 5 9). Discrimination tasks involving very dissimilar aldehydes were solved rapidly
by all subjects (100% correct after three to four trials), while when more similar odors were presented, the rats did not reach maximal performance until a much higher number of trials (80% correct at the eighth trial). We have shown that odor pairs differing by only one carbon are more difficult to discriminate from one another than odor pairs differing by two, three, or four carbons. The question remains whether this degree of difficulty correlates with the differences in the number of carbons per se, regardless of the absolute number of carbons in either molecule. We addressed this question by correlating all possible pairs of task acquisition curves (from Fig. 3A) with one another. The resulting correlation coefficients were then grouped by the difference in the difficulties of the two task acquisition curves that were correlated; for example, the two tasks “[4]CHO versus [3]CHO” and “[4]CHO versus [5]CHO” have the same difficulty because the difference in chain length between the two odorants is one in each case. The two tasks “[4]CHO versus [3]CHO” and “[4]CHO versus [8]CHO,” however, are of very different difficulty, because in one task the difference in carbon chain length between the two odorants is one, and in the other task it is four. The results (Fig. 3B) indicate that pairs of task acquisition curves derived from tasks of similar difficulty are more correlated with one another than are pairs of curves derived from tasks of dissimilar difficulty. For example, the acquisition curve “[3]CHO versus [4]CHO” was highly correlated (0.8) to the curve “[4]CHO versus [5]CHO” and the two curves were not significantly different (p . 0.2, paired t-test), while the correlation between the
FIG. 2. Generalization between aliphatic aldehydes of varying chain lengths. (A) The graph shows the average digging times [all rats (n 5 9) and all odor combinations (n 5 30)] in response to the conditioned odor (A), the test odor (B), and the control odor (X) as a function of the difference in chain lengths between the conditioned odor and the test odor. (B)The graph shows the average response levels [all rats (n 5 9)] to all test odors after conditioning with all six training odors ([3]CHO–[8]CHO). We compared the average response level to each test odor to the response to the control odor (two-way analysis of variance, *indicates p , 0.01).
ALIPHATIC ALDEHYDES
FIG. 3. Discrimination between aliphatic aldehydes of different chain length. (A) The graph shows the average acquisition curves for discrimination between [4]CHO and all other odorants used. (B) Correlation coefficients for the two by two comparisons of the acquisition curves shown in A. All the correlation coefficients are given. C(4 versus 3, 4 versus 5) is the correlation coefficient calculated for the task acquisition curves [4]CHO versus [4]CHO and [4]CHO versus [5]CHO. The correlation coefficients are grouped by the difference in difficulty of the two task acquisition curves. For example, [4] versus [6] and [4] versus [7] are relatively similar tasks (in one, the difference in chain length between the two odorants is 2; in the other, the difference in chain length is 3); whereas [4] versus [5] and [4] versus [8] are dissimilar tasks (in one, the difference in chain length between the two odorants is 1, and in the other it is 4). *Indicates when the two compared curves are significantly different from each other (p , 0.05, paired t-test).
501
502
LINSTER AND HASSELMO
curve “[4]CHO versus [5]CHO” and the curve “[4]CHO versus [7]CHO” was lower (0.5) and the two curves were significantly different (p , 0.01, paired t-test). Similarly, in the case of two “difficult tasks,” the acquisition curve “[6]CHO versus [8]CHO” was highly correlated to the curve “[7]CHO versus [8]CHO” and the curves were not statistically different (p . 0.2, paired t-test). All discrimination tasks could be learned by the rats given enough trials (after 10 trials, performance in all tasks increases to 100% correct responses). Thus, even “similar” odorants can be discriminated given appropriate motivation and acquisition time. The acquisition of the discrimination task could potentially have been influenced by the previous exposure to the aldehydes during the generalization task, as the same rats were used in both tasks. However, all the acquisition curves would have been affected similarly and the preexposure would be unlikely to influence the comparative results presented here. DISCUSSION
The results from the behavioral experiments presented herein suggest that the perceptual similarity of odorants can be predicted from the similarities in the neural responses evoked by these odorants in the olfactory bulb. Mitral cells in the olfactory bulb that are activated by aldehydes (among other odorants) tend to respond with increased firing rates to a range of aliphatic aldehydes differing by one or two carbons from the “best” odor (3,10,13). Similarly, in our experiments, rats generalized to aldehydes differing by one or two carbons from the conditioned odor. Thus, the large overlap in neural responses to these odorants as seen in the olfactory bulb is
predictive of a similar overlap in perception. The response to the conditioned odor is always at least threefold that of the test odors, suggesting that the overlap in representation leads to associative spread. This associative spread is behaviorally expressed as “confusion” between odorants. The fact that all rats learned to discriminate between similar aldehydes very rapidly (over a maximum of 10 trials) in the second experiment shows that, despite the generalization, the representations of these odorants are sufficiently different from each other to allow discrimination. Olfactory sensory neurons, presynaptic to mitral cells, respond to a larger number of compounds within a group of similar odorants than do individual mitral cells in the olfactory bulb (11). The relatively narrow response spectra of individual mitral cells to similar odorants within a group of related odorants is at least partly shaped by the inhibitory circuits in the olfactory bulb: when inhibition in the olfactory bulb is suppressed, mitral cells respond with increased firing rates to a larger number of compounds (13). We propose that generalization between similar components could arise from the overlap in responsiveness of mitral cells to these odorants, corresponding to a spread of association. In contrast, given appropriate motivation, learned discrimination between similar odorants necessitates the enhancement of the differences between the two patterns. Such an enhancement can arise from modulation of inhibition in the olfactory bulb during the learning task (5,6), or could arise from a self-organizing process in olfactory cortical areas. ACKNOWLEDGEMENTS
The authors thank Brian Smith for fruitful discussions and Thomas Cleland for helpful suggestions on the manuscript.
REFERENCES 1. Bunsey, M.; Eichenbaum, H.: Conservation of hippocampal memory function in rats and humans. Nature 379:255–257; 1996. 2. Cometto-Muniz, J. E.; Cain, W. S.; Abraham, M. H.: Nasal pungency and odor of homologous aldehydes and carboxylic acids. Exp. Brain Res. 118:180–188; 1998. 3. Imamura, K.; Mataga, N.; Mori, K.: Coding of odor molecules by mitral/tufted cells in rabbit olfactory bulb. I. Aliphatic compounds. J. Neurophysiol. 68:1986–2002; 1992. 4. Katoh, K.; Koshimoto, H.; Tani, A.; Mori, K.: Coding of odor molecules by mitral/tufted cells in rabbit olfactory bulb. II. Aromatic compounds. J. Neurophysiol. 70:2161–2175; 1993. 5. Linster, C.; Gervais, R.: Investigation of the role of interneurons and their modulation by centrifugal fibers in a neural model of the olfactory bulb. J. Comp. Neurosci. 3:225–246; 1996. 6. Linster, C.; Hasselmo, M.: Modulation of inhibition in an olfactory bulb model reduces overlap in the neural representation of olfactory stimuli. Behav. Brain Res. 84:117–127; 1997. 7. Linster, C.; Smith, B. H.: Generalization between binary odor mixtures and their components in the rat. Physiology and Behavior (in press).
8. Linster, C.; Smith, B. E.; Hasselmo, M. E.: Rats generalize between binary odor mixtures and their components. Am Chem Soc Meeting, Sarasota, FL; 1998. 9. Mori, K.; Mataga, N.; Imamura, K.: Differential specificities of single mitral cells in rabbit olfactory bulb for a homologous series of fatty acid odor molecules. J. Neurophysiol. 67:786–789; 1992. 10. Motokizawa, F.: Odor representation and discrimination in mitral/tufted cells of the rat olfactory bulb. Exp. Brain Res. 112:24–34; 1996. 11. Sato, T.; Hirono, J.; Tonoike, M.; Takebayashi, M.: Tuning specificities to aliphatic odorants in mouse olfactory receptor neurons and their local distribution. J. Neurophysiol. 72:2980–2989; 1994. 12. Scott, J. W.; Davis, L. M.; Shannon, D.; Kalan, C.: Relation of chemical structure to spatial distribution of sensory responses in the rat olfactory epithelium. J. Neurophysiol. 75:2036–2048; 1996. 13. Yokoi, M.; Mori, K.; Shigetada, N.: Refinement of odor molecule tuning by dendrodendritic synaptic inhibition in the olfactory bulb. Proc. Natl. Acad. Sci. USA 92:3371–3375; 1995.