The Journal of Experimental Biology 200, 2045Ð2055 (1997) 2045 Printed in Great Britain © The Company of Biologists Limited 1997 JEB0825

THE OLFACTORY MEMORY OF THE HONEYBEE APIS MELLIFERA

III. BILATERAL SENSORY INPUT IS NECESSARY FOR INDUCTION AND EXPRESSION OF OLFACTORY BLOCKING

ROBERT S. THORN AND BRIAN H. SMITH* Department of Entomology, 1735 Neil Avenue, Ohio State University, Columbus, OH 43210, USA

Accepted 7 May 1997

Summary The associative learning phenomenon termed ‘blocking’ than the antennae must be crucial for establishing demonstrates that animals do not necessarily associate a blocking. Further experiments show that this bilateral conditioned stimulus (e.g. X) with reinforcement if X is interaction between brain hemispheres is crucial during coincident with a second conditioned stimulus (e.g. A) that both the induction and the expression of blocking. This had already been associated with the same reinforcement. result implies that blocking involves an active inhibition of Blocking therefore represents a tactic that animals can use odor association and recall, and that this inhibition is to modulate associative learning in order to focus on the mediated by a structure that spans both brain hemispheres. most predictive stimuli at the expense of novel ones. Using This interpretation is consistent with a role for identified an olfactory blocking paradigm in the honeybee, we bilateral modulatory neurons in the production of investigated the mechanistic basis for olfactory blocking. blocking. We show that removing input from one antenna eliminates the blocking of one odor by another. Since antennal sensory neurons only project to the ipsilateral antennal lobe in the Key words: memory, honeybee, Apis mellifera, learning, olfactory honeybee, more central processing regions of the brain conditioning, blocking, odor mixtures.

Introduction Animals do not always ‘reflexively’ associate a conditioned reinforcement in a way that would produce robust associative stimulus (CS) with an unconditioned stimulus (US). In recent learning if A were not present. Blocking is a widespread years, theoretical accounts of learning have focused on several phenomenon and is central to understanding associative paradigms that demonstrate that animals can regulate whether learning of complex mixtures of stimuli. It was first described a CS enters into an association even though it may be properly by Kamin (1968, 1969) in rat associative conditioning, but has paired with reinforcement (Kamin, 1968, 1969; Rescorla and been found in a variety of other animals, including Wagner, 1972; Macintosh, 1974, 1983; Pearce and Hall, 1980; invertebrates such as honeybees (Smith and Cobey, 1994; Rescorla and Holland, 1982; Rescorla, 1988; Pearce, 1994). Smith, 1996, 1997) and slugs (Sahley et al. 1981), and This flexible association system probably reflects the need to blocking is even found in spinal reflexes (Illich et al. 1994). It extract pertinent sensory stimuli from a confusing, stimulus- may thus represent a basic and widespread tactic for rich environment every time that an animal learns an experience-dependent biasing of sensory processing. association (Smith, 1996). Since animals do not have unlimited Furthermore, blocking is independent of the type of US, sensory capacities, they have evolved strategies for focusing occurring in both appetitive (Kamin, 1968, 1969) and aversive sensory processing capacities on the most useful stimuli. In (Ross, 1985) conditioning, and, at least in the honeybee, it may most instances, this involves tuning sensory systems to stimuli be more robustly expressed among conditioned stimuli from that are predictive, and several central nervous system (CNS) the same sensory modality (Bitterman, 1996). tactics have evolved to facilitate this tuning. Little is known about the neuroanatomical substrates of One such tactic regards a learning phenomenon called blocking. Several studies of vertebrates have found that blocking, in which a CS that has been previously learned (e.g. hippocampal lesions disrupt blocking (Solomon, 1977; Rickett odor A) will substantially overshadow a second CS (e.g. odor et al. 1978), presumably by affecting memory of pretraining. X) that is added to it. The association of X with reinforcement Holland and Gallagher (1993a,b) have produced specific is largely blocked despite the fact that it is paired with blocking deficits with neuroanatomical lesions in the rat

*Author for correspondence (e-mail: [email protected]). 2046 R. S. THORN AND B. H. SMITH amygdala. They found that neurotoxic lesions in the central Conditioning protocols nucleus of the amygdala could affect the ability of changes in Honeybees were odor-trained using the restrained bee the US to unblock mixture training, and their interpretation was preparation as described elsewhere (Kuwabara, 1957; that the amygdala was important in directing attention to novel Bitterman et al. 1983; Menzel, 1990). In brief, individual CSÐUS pairs. These studies all suggest that blocking involves subjects were mounted in a harness that allowed them to move several distinct brain pathways of attention and memory to the their head, antennae and mouthparts. Each was trained to CS and the US. associate a brief 4 s pulse of odor in a moving airstream with It would be useful to compare this organization with that of a touch of 2.0 mol l−1 sucrose solution to the antenna. The other systems displaying blocking. Olfactory learning in the timing of odor delivery was controlled via computer. The honeybee (Apis mellifera) displays robust blocking, as sucrose acts as a US, releasing a motor pattern called the mentioned above, and is increasingly well-characterized at the proboscis extension reflex (PER) that bees use to ingest nectar. neuroanatomical level (for reviews, see Hammer and Menzel, PER was reinforced with a 0.4 µl drop of a 2.0 mol l−1 sucrose 1995; Menzel and Muller, 1996). In brief, odor-sensitive solution in each conditioning trial. The timing of US delivery antennal sensory cells project their axons into the antennal lobe was signaled to begin 3 s after odor onset and, given the time where they end in discrete glomeruli reminiscent of the it takes to consume the 0.4 µl droplet, would extend slightly vertebrate olfactory lobe. In the antennal lobe, they synapse beyond odor delivery. A subject that has learned the CSÐUS with different interneurons; most are local inhibitory association will frequently exhibit PER in response to the odor interneurons that project between glomeruli, but some project alone, or prior to US onset, after as few as 1Ð2 conditioning out to higher brain centers, specifically the mushroom bodies trials (Menzel, 1990). and lateral protocerebral lobes (Homberg, 1984; Flanagan and The conditioning protocol for odor blocking consisted of Mercer, 1989; Fonta et al. 1993). The former region is believed three phases of training conducted in parallel on two different to be an associative center (Erber et al. 1980; Mobbs, 1985; groups of honeybees using the procedure developed for Davis, 1993; de Belle and Heisenberg, 1994), while the function honeybees by Smith and Cobey (1994; Table 1). In the of the latter is less clear, although it is thought to be a premotor pretraining phase, subjects were conditioned in a cluster of four center. Odor blocking could occur at any or all of these levels. trials with a 10 min intertrial interval (ITI) (this interval was To investigate the potential involvement of these sites in the constant across all phases). All subjects experienced one of two production of blocking, it is particularly important to determine pure odorants as the CS during this phase. Group BLOCK was what sorts of experimental manipulations attenuate blocking. exposed to the blocking odorant (A), while the control group Such information is crucial for elucidating the neural and NOVEL received the other odorant (N). In the mixture phase, behavioral mechanisms of blocking in any system. In the each subject in both groups was conditioned with a mixture of present work, experimental manipulations were designed to two odorants as the new odor CS for a block of four trials. One affect the neural representation of the CS. We used proboscis of the odors was A and the other was a new odor X. This AX extension conditioning in honeybees specifically to examine compound odor was used to train both the groups NOVEL and how the antennae and the two hemispheres of the brain interact BLOCK; however, subjects in group BLOCK were the only in olfactory learning and blocking (Smith and Cobey, 1994; ones to have experienced one of the mixture components (i.e. Smith, 1996, 1997). We show that honeybees require A) during pretraining. A 0.4 µl droplet of 2.0 mol l−1 sucrose stimulation to both antennae to facilitate blocking and that the was again used as reinforcement in both phases. dynamics of this bilateral interaction are complex. In the final test phase, each subject was presented with a series of four consecutive unreinforced (extinction) trials consisting of exposure to a 4 s pulse of odorant X on its own. Materials and methods If pretraining significantly blocked acquisition of X during the Honeybee (Apis mellifera L.) workers were obtained either mixture phase, then subjects that had experienced A in the from indoor colonies (during the months of February and pretraining phase (i.e. group BLOCK; Table 1) would be March) or from colonies maintained outdoors (AprilÐMay). We expected to show significantly fewer PER responses on specifically used foragers for all of our training to minimize any average to X across the four trials in this phase than subjects variability due to age or behavioral caste. The indoor colonies that had experienced odor N (i.e. group NOVEL; Table 1). In were maintained in a flight room on a 16 h:8 h light:dark cycle and fed unscented sucrose solutions. ‘Foraging’ honeybees Table 1. Summary of treatment groups used in blocking from the indoor colonies typically flew towards the overhead experiments lights and briefly alighted on the net. Subjects from outdoor colonies were collected as they returned to their colonies from Treatment group Pretraining Mixture training Test foraging trips. Only pollen foragers were captured, and their NOVEL N → sucrose AX → sucrose X1, X2, X3, X4 pollen packets were analyzed to ascertain the type of flower that BLOCK A → sucrose AX → sucrose X1, X2, X3, X4 they had visited. Of the outdoor honeybees, only those that had foraged at the same type of flowers, as indicated by similar A, N and AX refer to odorants or a mixture. pollen types, were used and compared in each experiment. X1ÐX4 refer to four sequential extinction trials in the test phase. Bilateral sensory inputs and odor blocking 2047 all figures, we use line graphs to depict response probability test phase. Each subject was ranked on the basis of this sum across acquisition (reinforced) trials, whereas frequency (0Ð4 depending upon how often they demonstrated PER to the histograms depict the frequency of responses within a test odorant X). The rankings of subjects in the NOVEL and treatment population across the four test trials. BLOCK groups in each experiment were compared using The pure odorants geraniol and 1-hexanol were single-classification analysis of variance (ANOVA) for two counterbalanced as A and X throughout all experiments, groups (Sokal and Rohlf, 1981). The results from this test are although the results are separated to demonstrate that the reported as a F-statistics and always have one degree of identity of A or X did not qualitatively affect the conclusions. freedom. A Wilcoxon non-parametric test was also performed 2-Octanone was always used as N. For a generalization test, 1- in parallel to each ANOVA, which confirmed the significance octanol was used because it elicits stronger generalization of the F-statistic in every case. Statistical analyses of responses from subjects conditioned to 1-hexanol (Smith and generalization results were performed on the response trials to Menzel, 1989). Geraniol, 1-hexanol and 2-octanone have each the three different odorants and are compared using χ2-tests. been found to stimulate different classes of antennal receptors (Vareschi, 1971). Only geraniol has been identified as a component of honeybee pheromones: it is a part of the Results Nasonov gland pheromone (Pickett et al. 1980). 1-Hexanol and Using a protocol developed by Smith and Cobey (1994), we 1-octanol are found in some flower aromas (Knudsen et al. were able reliably to reproduce blocking of odor learning in 1993). To make up odor cartridges, 3 µl of pure odorant was honeybees (Fig. 1). Subjects were trained with the PER placed onto a strip of filter paper, which was then inserted into procedure to associate the odorant 1-hexanol (group BLOCK) a 1 ml glass syringe (see Smith, 1997, for details of odor or 2-octanone (group NOVEL) with sucrose reinforcement delivery). In the case of mixtures, 1.5 µl of each odorant in the during pretraining. Both groups received identical conditioning mixture was placed onto the filter paper strip. to the AX mixture, which contained geraniol and 1-hexanol. The training regimen to assess stimulus generalization Groups show patterns of acquisition during both phases that consisted of a block of five consecutive training trials with a are typical of blocking experiments (Smith and Cobey, 1994; 10 min ITI followed by a wait of 1 h, after which subjects were Smith, 1996). During the mixture phase, group BLOCK, which tested with a pseudorandomized series of three odors. The received pretraining with odor A (1-hexanol; Table 1), shows three test odorants included the conditioned odor (CS), an strong generalization to the mixture on the first mixture odor similar to the conditioned odor and a dissimilar odor. In training trial. Group NOVEL generalizes to a lesser degree but all trials, subjects were conditioned to either 1-hexanol or 1- reaches approximately the same asymptotic level of octanol, then tested with 1-hexanol, 1-octanol and geraniol. responding by trial 3 of mixture training. Previous experiments have shown that animals display a All subjects were then identically tested with geraniol, decreasing gradient of generalization across these odors, which was odor X in this experiment (Fig. 1B; Table 1). As responding best to their conditioned odor, less to the similar can be seen in the test responses, subjects that were initially odor and least to the dissimilar odor (Smith and Menzel, trained to 1-hexanol in group BLOCK showed on average 1989). fewer responses to geraniol in the test series than did subjects in group NOVEL (F=6.4, P<0.05, d.f.=1,53). That is, the Antennal manipulations frequency distribution of the response classes (0Ð4) is shifted To interfere with processing of odors in the antennal system, to the left in group BLOCK (open columns) relative to NOVEL we disrupted the processing capability of one antenna by either (hatched columns). Therefore, pretraining with 1-hexanol covering it or removing it. Removal consisted of cutting off blocked, or at least significantly hindered, complete association the antenna at the base and allowing the wound to heal for at of reinforcement with geraniol during mixture training. Note least 24 h. To cover an antenna, a small polyethylene tube that in this experiment the identities of A and X were actually (inner diameter 0.25 mm, outer diameter 0.76 mm) was cut to counterbalanced (see Fig. 2A,B below), but are reported here a length just longer than an antenna and was then slid over the for one pair only in order to demonstrate the blocking effect. target antenna. The open end was sealed with Vaseline or warm The rest of the Results section is organized into two sealing wax, while the end at the base of the antenna was sealed subsections. In the first, we present evidence that removing with Vaseline. This ‘sleeve’ did not physically harm the input from one antenna prevents or at least attenuates blocking. antenna and could subsequently be removed at different stages In the second subsection, we show that bilateral input is crucial of the experiment. Experiments in which the sleeve was at several different stages of the blocking process and that it switched between antennae demonstrated that the sleeve was extends to some other facets of odor learning. a very effective barrier to odors. Sleeves were discarded after a single use. Attenuation of blocking by removal of bilateral input As expected, when both antennae were left exposed, Statistical analyses blocking was evident regardless of whether geraniol (Fig. 2A; Statistical analyses were performed on the number of F=24.9, P<0.01, d.f.=1,28) or 1-hexanol (Fig. 2B; same data responses across the four extinction trials performed during the and statistics as in Fig. 1) was used as odor A (Table 1) during 2048 R. S. THORN AND B. H. SMITH

Fig. 1. Standard test for odor blocking. (A) Acquisition BLOCK NOVEL BLOCK NOVEL curves (lines) and (B) extinction tests (columns) for two 100 A 0.6 B sets of subjects trained in parallel in a BLOCK (odorant A; 90 P<0.05 1-hexanol, N=30) or NOVEL (odorant N; 2-octanone, 80 0.5 70 N=25) group (see Table 1). Each set was pretrained for four 0.4 trials. Subjects in both groups were equivalently 60 50 conditioned to a mixture of 1-hexanol and geraniol as AX 0.3 40 for an additional four trials. Acquisition curves always 30 Frequency 0.2 display the percentage of subjects in a treatment group that 20 proboscis extension Percentage showing 0.1 responded on a given trial. (B) After mixture training, 10 subjects were tested with odorant X (in this case geraniol) 0 0 across four extinction trials. Graphs of extinction responses 1 2 3 4 1 2 3 4 0 1 2 3 4 present the frequency of subjects that fall into the summed Pretraining Mixture training Number of responses response categories. For example, subjects that responded Trial number on all four trials fall into category 4; in this case, 10/25 NOVEL pretrained animals (40 % of total) responded in all four test trials, while 4/30 BLOCK pretrained animals responded (13 %). Stronger acquisition of odorant X during mixture training would be indicated by a shift in this frequency distribution to the right. A drawing of a head with antennae is used to indicate the antennal treatment that both the NOVEL and BLOCK groups received; in this case, that the antennae were left untreated (i.e. not covered or removed). The level of significance (P) or lack of significance (NS) is indicated on the figure.

pretraining. In contrast, when one antenna is either removed trials (Fig. 3A; F=3.7, P=0.05, d.f.=1,25). In comparison, for (Fig. 2C, F=0.1, NS, d.f.=1,24; Fig. 2D, F=0.05, NS, groups NOVEL and BLOCK in which subjects had one d.f.=1,31) or completely covered with an odor-impermeable antenna covered throughout training and testing, the blocking plastic sleeve (Fig. 2E, F=0.3, NS, d.f.=1,28; Fig. 2F, F=0.9, effect was attenuated and not significant (Fig. 3B; F=0.2, NS, NS, d.f.=1,38), subjects in groups BLOCK and NOVEL d.f.=1,30). Therefore, it is unlikely that covering one antenna learned to recognize odorant X equally well. In the last four cases, the lack of significance is not due simply to sample size. There were still no significant differences between BLOCK BLOCK NOVEL 0.6 A B and NOVEL when data from Fig. 2C,D were combined P<0.01 P<0.05 (F=0.2, NS, d.f.=1,57) or when data from Fig. 2E,F were 0.5 combined (F=1.1, NS, d.f.=1,68). Since cutting and covering 0.4 the antenna were indistinguishable in their effects on blocking, antennal covering was used in all subsequent experiments. 0.3 It is conceivable that subjects with only a single usable 0.2 antenna displayed reduced blocking because of an impairment 0.1 in perception of the pretraining odorant (i.e. A or N). Therefore, in a subsequent experiment, the length of 0 NS pretraining was doubled to eight trials. However, blocking was 0.6 C NS D still attenuated in this experiment (Fig. 3). For groups NOVEL 0.5 and BLOCK in which subjects did not have one antenna 0.4 covered, the normal blocking effect was evident during test 0.3

Fig. 2. Effects of unilateral antennal removal or covering on odor Frequency 0.2 blocking. Extinction tests for independent groups of subjects that were 0.1 trained in BLOCK or NOVEL groups with intact antennae (A,B), with 0 one antenna removed 1 day prior to training (C,D) or with one antenna F covered 2 h prior to pretraining and throughout all training and testing 0.6 E NS NS (E,F). Each pair of graphs represents counterbalanced odor 0.5 treatments; that is, the left-hand column (A,C,E) used geraniol as the pretraining odorant A and 1-hexanol as X, whereas the right-hand 0.4 column (B,D,F) used the reverse. For C, D, E and F (and indeed all 0.3 subsequent graphs), the antenna that received treatment (i.e. left or 0.2 right) was counterbalanced across individuals, although for convention we always display the left antenna as representative of 0.1 treatment. Sample sizes for groups NOVEL and BLOCK are, 0 respectively: (A) 16, 14; (B) 25, 30; (C) 14, 12; (D) 12, 21; (E) 12, 0 1 2 3 4 0 1 2 3 4 18; (F) 15, 25. Number of responses Bilateral sensory inputs and odor blocking 2049

BLOCK NOVEL BLOCK NOVEL B 0.6 A B 0.6 A P<0.01 NS P<0.05 NS 0.5 0.5 0.4 0.4 0.3 0.3

Frequency 0.2 0.2 0.1 0.1 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0.6 C P<0.01 Number of responses Number of responses Frequency 0.5 Fig. 3. Effects of extended pretraining on odor blocking. Extinction tests for independent sets of animals that were conditioned in BLOCK 0.4 or NOVEL groups as in Fig. 1, but with pretraining consisting of eight 0.3 trials. (A) Extinction responses in subjects with both antennae uncovered. (B) Extinction responses from NOVEL and BLOCK 0.2 groups in which subjects had one antenna covered throughout all 0.1 training and testing periods. Sample sizes for groups NOVEL and 0 BLOCK are, respectively: (A) 13, 14; (B) 16, 16. 0 1 2 3 4 Number of responses slowed the ability to learn the characteristics of odor A during Fig. 4. Effects of covering one whole or two half-antennae upon odor pretraining, at least in as much as it would have impaired its blocking. Extinction tests for subject that had both antenna exposed ability to block X during mixture training. (A), one antenna covered (B) or two basal half-antennae covered (C). Furthermore, attenuation of blocking was specific to Sample sizes for groups NOVEL and BLOCK are, respectively: manipulations that silenced one complete antenna, since (A) 23, 23; (B) 28, 30; (C) 25, 26. blocking occurred as normal when only half of each antenna was covered (Fig. 4). Groups NOVEL and BLOCK showed presented during the test phase. The reduced surface area of a the blocking effect when both antennae were exposed single antenna could fail to provide enough information about (Fig. 4A; F=10.6, P<0.01, d.f.=1,44) and when both distal the pretraining odor so that the animal generalized too broadly halves were exposed (Fig. 4C; F=7.9, P<0.01, d.f.=1,49), but between the odors A and X. That is, X might be perceived as not when one complete antenna was covered (Fig. 4B; F=0.8, more similar to A by single-antenna subjects than by subjects NS, d.f.=1,56). This manipulation resulted in comparable areas able to use both antennae. However, generalization tests do not of exposed sensory epithelium in subjects with two half- necessarily support this explanation (Fig. 5). In those tests, antennae and those with one whole antenna. The result generalization to geraniol, which was the dissimilar odor, specifically suggests that it is the loss of bilateral input that is provided a gauge of generalization between odors used as A crucial and not just the general decline in sensory input levels. and X in blocking experiments (i.e. 1-hexanol and 1-octanol Finally, attenuation of blocking in subjects with a single were training odorants). In this case, generalization was not antenna could have been due to a reduced discrimination of X different across the three antennal treatment groups (dissimilar:

Both antennae exposed, N=51 One antenna covered, N=52 Distal halves exposed, N=61 100 A B Fig. 5. Effects of covering one whole or two half-antennae 90 on acquisition and generalization. (A) Half of the subjects 80 in each group were conditioned to 1-hexanol and the 70 remaining half to 1-octanol. (B) All were then tested in a 60 randomized series of three trials, once each with the 50 conditioned odor, the similar odor (i.e. 1-octanol for 1- 40 hexanol-conditioned subjects and vice versa) and with extension geraniol. Given that only one extinction trial was 30 performed with each odor, extinction responses are 20 presented as the percentage of subjects that responded on 10 Percentage showing proboscis each of these three trials. The significance levels indicated 0 in the graph reflect a comparison of the three antennal 1 2 3 4 5 Training Similar Dissimilar treatments within each test condition. Trial number Test odor 2050 R. S. THORN AND B. H. SMITH

BLOCK NOVEL BLOCK NOVEL 100 A 0.6 B NS 90 80 0.5 70 0.4 60 50 0.3 40 30 0.2 20 0.1 10 0 0 Fig. 6. Effects of switching the cover from one antenna to the 100 C 0.6 D NS other between pretraining and mixture training. (A) Acquisition 90 Frequency trials and (B) test trials of NOVEL and BLOCK groups that had 80 0.5 70 the same antenna covered. In contrast to other covered groups, 0.4 however, in this group the cover was removed and replaced 60 50 over the same antenna between the pretraining and mixture 0.3 40 training phases, leaving the antenna covered during the mixture Percentage showing proboscis extension 30 0.2 training and test phases. Thus, it serves as the control for the 20 0.1 group that experienced a switch. (C,D) The same data for 10 subjects that had the cover switched between the two antennae 0 0 between pretraining and mixture training. Sample sizes for 1 2 3 4 1 2 3 4 0 1 2 3 4 groups NOVEL and BLOCK are, respectively: (A,B) 18, 20; Pretraining Mixture training Number of responses (C,D) 27, 27. Trial number

χ2=2.2, NS). Response levels to the CS itself did not differ NOVEL odorant to AX is affected by the switch in covering across the three groups (training: χ2=0.6, NS). the antennae (χ2=6.9, P<0.01). These results support earlier In contrast to these two test conditions, the antenna-covered reports that information about A is not shared between the two group did tend to generalize more in the similar test condition antennal lobes even when one is covered and unstimulated (i.e. 1-hexanol was tested in groups conditioned to 1-octanol, (Masuhr and Menzel, 1972; Mercer and Menzel, 1982). The and vice-versa). The test statistic in this case was barely non- depression of AX learning during the initial trials of mixture significant (similar: χ2=5.8, 0.1>P>0.05). Thus, covering one training in animals with the cover switched between antennae antenna has a slight affect on the ability to differentiate similar further suggests that the US presentation alone could act to odors, but this is not likely to have had a significant impact build up a resistance to learning ‘new’ odors using the naive upon our blocking experiments, in which we used dissimilar antenna. This pattern defines a classic USÐpre-exposure effect, training and test odors. which has been documented in honeybee odor training (Abramson and Bitterman, 1986). Bilateral input is specifically necessary during testing It was thus important to know when both antennae were It is possible that the attenuation effect caused by covering necessary for blocking. To dissect out this sensitive period, one an antenna might be restricted to covering during one of the antenna was covered or left exposed during different phases of three phases of the training procedure. Therefore, several training and testing in five groups of subjects (Fig. 7). experiments were performed during which antennal covers Blocking was attenuated if the antenna was covered during were moved or removed during one or two of the three training pretraining (Fig. 7B; F=0.2, NS, d.f.=1,34), but was still phases (see Table 1). evident if the antenna was covered only during mixture training If the cover is briefly removed and then replaced over the (Fig. 7C; F=7.1, P<0.05, d.f.=1,26). Covering an antenna only same antenna between the pretraining and mixture training during the test phase also prevented blocking (Fig. 7D; F=0.1, phases, blocking still fails to occur (Fig. 6B; F=0.7, NS, NS, d.f.=1,42). Covering an antenna during pretraining and d.f.=1,36). Blocking is still attenuated if different antennae are mixture training with subsequent uncovering during the test covered during pretraining and mixture training/testing phase did not attenuate blocking (Fig. 7E; F=5.8, P<0.05, (Fig. 6D; F=0.7, NS, d.f.=1,52). Interestingly, under these d.f.=1,43), but covering during the mixture training and test conditions, subjects in group BLOCK do not show as much phases did attenuate blocking (Fig. 7F; F=2.0, NS, d.f.=1,31). generalization from A to AX on the first mixture training trial Taken together, these results suggest that bilateral antennal (Fig. 6C) as do subjects with one antennae covered (Fig. 6A) input is essential during the test phase for blocking to occur, or with both antennae exposed (Fig. 1A). A comparison of the but its absence during other phases does not always attenuate response probability to AX on trial one of mixture training blocking. This pattern of results is perplexing and suggests that between group BLOCK in Fig. 6A,C yields a significant blocking may involve a complex interaction between brain difference (χ2=49.6, P<0.001). Even generalization from the hemispheres, an idea that we will return to in the Discussion. Bilateral sensory inputs and odor blocking 2051

BLOCK NOVEL possible that the structures mediating odor blocking span the 0.6 A B hemispheres at this level. These results must now be placed into the broader context 0.5 of their meaning for olfactory blocking. One consideration 0.4 P<0.05 NS concerns the ecological role of a phenomenon such as 0.3 blocking. Smith (1996) has argued that blocking acts in a sense like a filter, such that odors that are less reliably present in a 0.2 floral bouquet might be less well represented in the memory 0.1 (and thus blocked more readily). Why should this phenomenon 0 of inter-hemispheric inhibition be important? The answer must await further research, but it might play a role in the detection 0.6 C D of spatially restricted odor gradients between antennae, as 0.5 described by Kramer (1976). The mechanism we describe here 0.4 P<0.05 NS might function to enhance the signal-to-noise ratio embedded 0.3 in such small-scale gradients, perhaps by highlighting odors present at different concentrations at each antenna during odor Frequency 0.2 learning. To answer this question, we must consider the 0.1 mechanism for olfactory blocking, towards which we now 0 orient the Discussion. 0.6 E F Blocking is a phenomenon of the central nervous system 0.5 Blocking and related processes in vertebrates are localized NS 0.4 P<0.05 to specific regions of the brain; for example, the hippocampus 0.3 and amygdala (Holland and Gallagher, 1993a,b; Han et al. 1995; Hatfield et al. 1996). Indeed, most higher-order 0.2 information processing in vertebrates occurs beyond the 0.1 sensory epithelium. However, recent work on odor and odor 0 mixture processing in such invertebrates as lobsters (Derby et 0 1 2 3 4 0 1 2 3 4 al. 1994), honeybees (Getz and Akers, 1994; Bhagavan and Number of responses Smith, 1997) and cockroaches (Getz and Akers, 1996) suggests that several non-linear interaction effects in odor coding can Fig. 7. Effects of covering one antenna at different times during the occur in the sensory cells of the antenna. Other work on pretraining, mixture training and testing. The three honeybee diagrams in each figure section represent antennal treatment for honeybees suggests that neighboring sensory cells may NOVEL and BLOCK groups during each of the three phases. influence one another (Akers and Getz, 1993; Getz and Akers, Antennae were covered as follows: (A) not covered; (B) pretraining 1994), and the intercellular messenger nitric oxide may be only; (C) mixture training only; (D) testing only; (E) pretraining and capable of mediating this interaction (Breer and Shephard, mixture training; (F) mixture training and testing. Note how the right- 1993). Therefore, it must be considered possible that blocking hand column of experiments (B,D,F) shows attenuation of blocking, in olfactory mixtures could occur at the level of the sensory while the left-hand column (A,C,E) does not. Sample sizes for groups epithelium in the antennae. NOVEL and BLOCK are, respectively: (A) 14, 9; (B) 16, 20; (C) 12, The present results argue against such a peripheral site for 16; (D) 15, 29; (E) 16, 29; (F) 15, 18. blocking and support other lines of evidence that olfactory blocking does not arise in the sensory cells of the antennal Discussion system (Smith and Cobey, 1994; Smith, 1996). Since covering We have used the proboscis extension reflex in the one antenna, but not two half-antennae, can attenuate blocking, honeybee (Bitterman et al. 1983; Menzel, 1990) to explore it seems unlikely that the antennal sensillae contribute how the honeybee brain might be organized to accomplish substantially to blocking. If the sensillae played a significant blocking (Smith and Cobey, 1994). Perhaps the most role in inducing blocking, then blocking should occur important overall finding is that bilateral input is necessary for regardless of whether one whole antenna or two half-antennae one odor to block a second. This result suggests that the two are covered. Furthermore, Smith and Cobey (1994) have hemispheres of the honeybee brain interact in the production shown that other kinds of presentation of odor A during of blocking, even though excitatory information about a pretraining do not produce blocking; that is, the blocking effect learned odorant does not transfer between the hemispheres at is limited to forward pairing of odor A with reinforcement the level of the antenna or antennal lobe (Masuhr and Menzel, during pretraining. This result is consistent with blocking due 1972; Fig. 6A,C). Since cooling experiments by Erber et al. to associative mechanisms (Pearce and Hall, 1980; Rescorla, (1980) show that this type of information can be shared 1988) but not to non-associative mechanisms (Smith, 1996). between mushroom bodies of different hemispheres, it is Therefore, the search for the physiological mechanisms 2052 R. S. THORN AND B. H. SMITH underlying blocking must turn to more central areas of the occluded could only remember the training odor association brain and account for several characteristics revealed in our through the naris that was open at the time of training present analyses. (Kucharski et al. 1986). This odor memory could be accessed by the contralateral olfactory bulb after the commissure had Olfactory blocking must involve inter-hemispheric developed, indicating that both sides of the olfactory bulb were communication monitored for memories in the mature animal (Kucharski and The need for interhemispheric communication is surprising, Hall, 1988). This delayed bilateralization may have some particularly since an antenna, its associated antennal lobe and important consequences for neonatal rats, since unilateral odor mushroom bodies (collectively the antennal system) appear to learning in bees has been shown to be important for navigating identify odors autonomously from the contralateral antennal along an odor gradient (Martin, 1964; Kramer, 1976). Given system (Masuhr and Menzel, 1972; Fig. 6A,C). Several studies that similar odor navigation is found in some reptiles have documented how olfactory memories formed in one (Schwenk, 1994) and may be important in neonatal marsupials antennal system are not accessible to the contralateral (Gemmell and Rose, 1989), it is conceivable that it could also hemisphere via the other antenna. If subjects are conditioned be used by blind neonatal rats. with one antenna exposed while the other is covered, the The results of this paper strongly suggest that input from odorÐsucrose association is restricted unilaterally to the trained both antennae is crucial for initiating blocking in honeybees. antennal pathway. If the cover is switched to the other antenna While odor identity may be a product of each autonomous after training, the animal fails to respond to the conditioned hemisphere (Masuhr and Menzel, 1972; Fig. 6), it is apparent odor and must be retrained in order to obtain a conditioned from our results that inputs from both antennae are integrated response (Masuhr and Menzel, 1972; Macmillan and Mercer, when utilizing odor information in higher-order decisions. 1987). There are no direct decussations between contralateral Anatomically, the connections between the brain hemispheres antennal sensory cells and the antennal lobes, so this that help mediate blocking might pass through the mushroom phenomenon has traditionally been interpreted as indicating bodies, since that is one of the points at which olfactory that odor identity in associative learning is specified in the information decussates. Mushroom bodies clearly have a ipsilateral antennal lobe and/or the associated ipsilateral critical role in associative odor learning in Drosophila (de mushroom body. Belle and Heisenberg, 1994). In addition, there are There might, however, be some inhibitory interaction interneurons that span both pairs of mushroom bodies and between the hemispheres that would give rise to the lower antennal lobes (Hammer, 1993; Hammer and Menzel, 1995), generalization to the mixture (AX) and the subsequent delay so it is certainly possible that the interaction we describe may in acquisition we observed when the cover was switched not involve intrinsic mushroom body neurons. Such between antennae. This delay could be explained by the modulation by central associative areas also has a correlate in multiple US presentations without associated stimulation of the rats, where hippocampal lesions can interfere with an odor covered antenna during pretraining (i.e. a USÐpre-exposure learning task similar to blocking (Schmajuk et al. 1983) as well effect, as shown for honeybees by Abramson and Bitterman, as blocking in non-odor learning paradigms (Solomon, 1977; 1986). Any subsequent associational odor learning in that Rickert et al. 1978). antennal system would first have to overcome this inhibition. Note that our hypotheses do not eliminate a role for the Furthermore, any transfer of inhibition between antennae antennal lobes in odor blocking. The commissure that has been might be related to the mechanism that gives rise to blocking. described as passing between the two antennal lobes (Mobbs, This situation is rather different from that observed in rats, 1985) affords one avenue for direct cross-talk and it appears to where unilateral odor training results in bilateral memory for contain neuromodulatory processes. It is also possible that the CS odor (Teitelbaum, 1971). This difference is somewhat bilateral inputs are weighed in the mushroom bodies, and that surprising, given that in both honeybees and rats the laterality this information is then centrifugally relayed back to the of the left and right olfactory pathways is largely maintained antennal lobes where it aids in setting up blocking conditions. until higher-order associative centers are reached. Mammals Interneurons that connect the mushroom bodies and each appear to have no direct decussations between the left and right antennal lobe are present in the honeybee (Hammer, 1993). olfactory lobes, while honeybees have only a small Thus, information could flow between the antennal lobes or commissure (Mobbs, 1985). In rats, a portion of the anterior from the mushroom body to the antennal lobes via this route. commissure of the cerebrum must be lesioned in order to Similar centrifugal projections to the olfactory lobes in rats prevent memory sharing (Teitelbaum, 1971; Kucharski and have important effects upon odor learning and memory in Hall, 1988). Thus, mature rats appear to be capable of neonates (Wilson and Sullivan, 1994) and adults (Brennan et identifying odors unilaterally, but the information is typically al. 1990). quickly shared with the contralateral hemisphere. Fig. 8 summarizes how disrupting bilateral inputs attenuates A striking parallel with honeybees exists in the early blocking and reveals the potential complexity of the underlying postnatal development of rats, where prior to day 12 the mechanism. Disruption throughout all three phases interferes commissures between the olfactory association areas are not with blocking. This result would be consistent with a model that yet organized. Younger animals trained with one naris incorporates transmission of inhibitory information between the Bilateral sensory inputs and odor blocking 2053

Mixture This model assumes unilateral odor processing, but several Pretraining training Test Block? different components of it can be made ‘bilateral’ to test Yes whether the model can accommodate our results: (1) the excitatory input could have bilateral effects; (2) the inhibitory No interneurons could have bilateral effects; or (3) the US- Yes correlating interneuron(s) could have bilateral effects. It remains to be determined how or whether these modifications No to the model can accommodate our results, particularly those Yes presented in Fig. 7E. The answer to this question remains to be elucidated by future behavioral and physiological analyses. No

No Support for this study was provided by a grant from NIMH to B.H.S. No

No References ABRAMSON, C. I. AND BITTERMANN, M. E. (1986). The USÐpre- exposure effect in honeybees. Anim. Learn. Behav. 14, 374Ð379. Fig. 8. Diagram summarizing antennal covering configurations that AKERS, R. P. AND GETZ, W. M. (1993). Response of olfactory receptor allowed for or attenuated the blocking effect. Shaded half-bars show neurons in honeybees to odorants and their binary mixtures. J. when during training one antenna was covered. comp. Physiol. A 173, 169Ð185. BHAGAVAN, S. AND SMITH, B. H. (1997). Olfactory conditioning in the antennal lobes. That is, when activity in one hemisphere is honey bee, Apis mellifera: effects of CS intensity. Physiol. Behav. eliminated, then inhibition of the contralateral hemisphere (in press). would be diminished and would be incapable of giving rise to BITTERMAN, M. E. (1996). Comparative analysis of learning in blocking. Furthermore, blocking was attenuated by disruption honeybees. Anim. Learn. Behav. 24, 123Ð141. during the initial acquisition phase, by disruption during the BITTERMANN, M. E., MENZEL, R., FIETZ, A. AND SCHAEFER, S. (1983). testing phase and by disruption during both the mixture and the Classical conditioning of proboscis extension in honeybees (Apis testing phases. Such attenuation did not occur when bilateral mellifera). J. comp. Psychol. 97, 107Ð119. BREER, H. AND SHEPHERD, G. M. (1993). Implications of the input was disrupted only during mixture training. These results NO/cGMP system for olfaction. Trends Neurosci. 16, 5Ð9. are also consistent with a bilateral inhibition model and would BRENNAN, P., KABA, H. AND KEVERNE, E. B. (1990). Olfactory seem to indicate that bilateral inhibitory transmission is most recognition: A simple memory system. Science 250, 1223Ð1226. important during the acquisition and testing phases. However, DAVIS, R. L. (1993). Mushroom bodies and Drosophila learning. the experiment that disrupted bilateral input during the Neuron 11, 1Ð20. acquisition and mixture phases (Fig. 7E) revealed that the DE BELLE, J. S. AND HEISENBERG, M. (1994). Associative odor learning mechanism must be more complex than the one just implied. in Drosophila abolished by chemical ablation of mushroom bodies. That is, blocking remained intact in that experiment (Fig. 7E) Science 263, 692Ð695. even though disruption of bilateral inputs only during DERBY, C., BURGESS, M., OLSON, K., SIMON, T. AND LIVERMORE, A. acquisition was sufficient to attenuate it (Fig. 7B). (1994). Mechanisms of detection and discrimination of mixtures in Linster and Smith (1997) have developed a computational the olfactory system of spiny lobsters. In Olfaction and Taste XI (ed. K. Kurihara, N. Suzuki and H. Ogawa), pp. 775Ð777. Tokyo: model of antennal lobe glomeruli to examine how the antennal Springer-Verlag. lobes could be organized to produce odor identification and ERBER. J., MASUHR, T. AND MENZEL, R. (1980). Localization of short- blocking. The primary features of this model are (1) that odor- term memory in the brain of the bee, Apis mellifera. Physiol. stimulated sensory axons are active in a discrete subset of Entomol. 5, 343Ð358. antennal lobe glomeruli; (2) that local inhibitory interneurons FLANAGAN, D. AND MERCER, A. R. (1989). Morphology and response activated by these glomeruli act to suppress activity in other, characteristics of neurones in the deutocerebrum of the brain of the less-active glomeruli in the lobe; and (3) that a honeybee Apis mellifera. J. comp. Physiol. A 164, 483Ð494. neuromodulatory neuron active only during US presentation FONTA, C., SUN, X.-J. AND MASSON, C. (1993). Morphology and acts to stabilize this pattern of heightened synaptic activity for spatial distribution of bee antennal lobe interneurons responsive to the learned odors and depressed synaptic activity for all other odors. Chem. Senses 18, 101Ð119. odors. This mechanism can sharpen the antennal lobe GEMMELL, R. T. AND ROSE, R. W. (1989). The senses involved in movement of some newborn Macropodoidea and other marsupials representation of an odor by reducing anomalous background from cloaca to pouch. In Kangaroos, Wallabies and Rat Kangaroos activity (Linster and Masson, 1995). Furthermore, if (ed. P. Jarman, I. D. Hume and G. Grigg), pp. 339Ð347. Sydney: reinforcement enhances the spread of inhibition, it could also Surrey Beatty. act to temporarily disable the representation of another odor GETZ, W. M. AND AKERS, R. P. (1994). Honeybee olfactory sensilla (e.g. X) presented with a learned odor (Linster and Smith, behave as integrated processing units. Behav. neural Biol. 61, 1997), which produces blocking. 191Ð195. 2054 R. S. THORN AND B. H. SMITH

GETZ, W. M. AND AKERS, R. P. (1996). Response of American MACMILLAN, C. S. AND MERCER, A. R. (1980). An investigation of the cockroach (Periplaneta americana) olfactory receptors to selected role of dopamine in the antennal lobes of the honeybee, Apis alcohol odorants and their binary combinations. Chem. Senses (in mellifera. J. comp. Physiol. A 160, 359Ð366. press). MARTIN, H. (1964). Zur Nahorientierung der Biene im Duftfeld. HAMMER, M. (1993). An identified neuron mediates the unconditioned Zugleich ein Nachweis fur die Osmotropotaxis bei Insekten. Z. stimulus in associative olfactory learning in honeybees. Nature 366, vergl. Physiol. 48, 481Ð533. 59Ð63. MASUHR, T. AND MENZEL, R. (1972). Learning experiments on the use HAMMER, M. AND MENZEL, R. (1995). Learning and memory in the of side-specific information in the olfactory and visual system in honeybee. J. Neurosci. 15, 1617Ð1630. the honey bee (Apis mellifera). In Information Processing in the HAN, J.-S., GALLAGHER, M. AND HOLLAND, P. (1995). Hippocampal Visual Systems of Arthropods (ed. R. Wehner), pp. 315Ð321. lesions disrupt decrements but not increments in conditioned Berlin: Springer-Verlag. stimulus processing. J. Neurosci. 15, 7323Ð7329. MENZEL, R. (1990). Learning, memory and ‘cognition’ in honey bees. HOLLAND, P. C. (1984). Unblocking in Pavlovian appetitive In Neurobiology of Comparative Cognition (ed. R. P. Kesner and conditioning. J. exp. Psych.: Anim. Behav. Proc. 10, 476Ð497. D. A. Olton), pp. 237Ð292. Hillsdale, NJ: Lawrence Erlbaum HOLLAND, P. C. (1988). Excitation and inhibition in unblocking. J. Assoc. exp. Psych.: Anim. Behav. Proc. 14, 261Ð279. MENZEL, R. AND MÜLLER, U. (1996). Learning and memory in HOLLAND, P. C. AND GALLAGHER, M. (1993a). Effects of amygdala honeybees: From behavior to neural substrates. Ann. Rev. Neurosci. central nucleus lesions on blocking and unblocking. Behav. 19, 379Ð404. Neurosci. 107, 235Ð245. MERCER, A. R. AND MENZEL, R. (1982). The effects of biogenic HOLLAND, P. C. AND GALLAGHER, M. (1993b). Amygdala central amines on conditioned and unconditioned responses to olfactory nucleus lesions disrupt increments, but not decrements, in CS stimuli in the honeybee Apis mellifera. J. Comp. Physiol. A 145, processing. Behav. Neurosci. 107, 246Ð253. 363Ð368. HOMBERG, U. (1984). Processing of antennal information in extrinsic MOBBS, P. G. (1985). Brain structure. In Comprehensive Insect mushroom body neurons of the bee brain. J. comp. Physiol. A 154, Physiology, Biochemistry and Pharmacology, vol. V, Nervous 825Ð836. System: Structure and Function (ed. G. A. Kerkut and L. I. Gilbert), ILLICH, P. A., SALINAS, J. A. AND GRAU, J. W. (1994). Latent pp. 300Ð370. New York: Pergamon Press. inhibition, overshadowing and blocking of a conditioned PEARCE, J. M. (1994). Similarity and discrimination: A selective antinocioceptive response in spinalized rats. Behav. neural Biol. 62, review and a connectionist model. Psychol. Rev. 101, 587Ð607. 140Ð150. PEARCE, J. M. AND HALL, G. (1980). A model for Pavlovian learning: KAMIN, L. J. (1968). Attention-like processes in classical Variations in the effectiveness of conditioned but not of conditioning. Miami Symposium on the Prediction of Behaviors: unconditioned stimuli. Psychol. Rev. 87, 532Ð552. Aversive Stimulation (ed. M. R. Jones), pp. 9Ð32. Coral Gables, FL: PICKETT, J. A., WILLIAMS, I. H., MARTIN, A. P. AND SMITH, M. C. Miami University Press. (1980). Nasanov pheromone of the honeybee, Apis mellifera L. KAMIN, L. J. (1969). Predictability, surprise, attention and (Hymenoptera, Apidae). I. Chemical characterization. J. chem. conditioning. In Punishment and Aversive Behavior (ed. B. A. Ecol. 6, 425Ð434. Campbell and R. M. Church), pp. 279Ð296. New York: Appleton- RESCORLA, R. A. (1988). Behavioral studies of Pavlovian Century-Crofts. conditioning. A. Rev. Neurosci. 11, 329Ð352. KNUDSEN, J. T., TOLLSTEN, L. AND BERGSTROM, L. G. (1993). Floral RESCORLA, R. A. AND HOLLAND, P. C. (1982). Behavioral studies of scents Ð A checklist of volatile compounds isolated by head-space associative learning in animals. A. Rev. Psychol. 33, 265Ð308. techniques. Phytochem. 33, 253Ð280. RESCORLA, R. A. AND WAGNER, A. R. (1972). A theory of Pavlovian KRAMER, E. (1976). The orientation of walking honey bees in odour conditioning: Variations in the effectiveness of reinforcement and fields with small concentration gradients. Physiol. Ent. 1, 27Ð37. nonreinforcement. In Classical Conditioning II: Current Research KUCHARSKI, D. AND HALL, W. G. (1988). Developmental change in and Theory (ed. A. H. Black and W. F. Prokasy), pp. 64Ð99. New the access to olfactory memories. Behav. Neurosci. 102, 340Ð348. York: Appleton-Century-Crofts. KUCHARSKI, D., JOHANSON, I. B. AND HALL, W. G. (1986). Unilateral RICKERT, E. J., BENNETT, T. L., LANE, P. AND FRENCH, J. (1978). olfactory conditioning in 6-day-old rat pups. Behav. Neurosci. 46, Hippocampectomy and the attenuation of blocking. Behav. Biol. 22, 472Ð490. 147Ð160. KUWABARA, M. (1957). Bildung des bedingten Reflexes von Pavlovs ROSS, R. T. (1985). Blocking of conditioned analgesia. Learning Typus bei der Honigbiene, Apis mellifera. J. Fac. Sci. Hokkaido Motivation 16, 173Ð189. Univ. (Series 6) 13, 458Ð464. SAHLEY, C. L., RUDY, J. W. AND GELPERIN, A. (1981). An analysis of LINSTER, C. AND MASSON, C. (1995). A neural model of olfactory associative learning in a terrestrial mollusc. I. Higher-order sensory memory in the honeybee’s antennal lobe. Neur. Comput. conditioning, blocking and a transient US pre-exposure effect. J. 8, 94Ð114. comp. Psychol. 144, 1Ð8. LINSTER, C. AND SMITH, B. H. (1997). An investigation of SCHMAJUK, N. A., SPEAR, N. E. AND ISAACSON, R. L. (1983). Absence overshadowing, blocking and unblocking in binary odorant of overshadowing in rats with hippocampal lesions. Physiol. mixtures using a computational model of the honey bee antennal Psychol. 8, 254Ð261. lobe circuitry. Behav. Brain Res. (in press). SCHWENK, K. (1994). Why snakes have forked tongues. Science 263, MACINTOSH, N. J. (1974). The Psychology of Animal Learning. 1573Ð1577. London: Academic Press. SMITH, B. H. (1996). The role of attention in learning about odorants. MACINTOSH, N. J. (1983). Conditioning and Associative Learning. Biol. Bull. mar. biol. Lab., Woods Hole 191, 76Ð83. Oxford: Oxford University Press. SMITH, B. H. (1997). An analysis of blocking in binary odorant Bilateral sensory inputs and odor blocking 2055

mixtures: An increase but not a decrease in intensity of SOLOMON, P. (1977). Role of the hippocampus in blocking and reinforcement produces unblocking. Behav. Neurosci. 111, conditioned inhibition of the rabbit’s nictitating membrane 57Ð69. response. J. comp. physiol. Psychol. 91, 407Ð417. SMITH, B. H. AND COBEY, S. (1994). The olfactory memory of the TEITELBAUM, H. (1971). Lateralization of olfactory memory in the honeybee Apis mellifera. II. Blocking between odorant in binary split-brain rat. J. comp. physiol. Psychol. 76, 51Ð56. mixtures. J. exp. Biol. 195, 91Ð108. VARESCHI, E. (1971). Duftunterscheidung bei der Honigbiene Ð SMITH, B. H. AND MENZEL, R. (1989). The use of electromyogram Einzell-Ableitungen und Verhaltensreaktionen. Z. vergl. Physiol. recordings to quantify odorant discrimination in the honey bee, 75, 143Ð173. Apis mellifera. J. Insect Physiol. 35, 369Ð375. WILSON, D. A. AND SULLIVAN, R. M. (1994). Neurobiology of SOKAL, R. R. AND ROHLF, F. J. (1981). Biometry. 2nd edition. New associative learning in the neonate: early olfactory learning. Behav. York: W. H. Freeman and Company. neural Biol. 61, 1Ð18.