Brain, Vol. 125, No. 1, 86-101,
January 1, 2002
© 2002 Oxford University Press
Olfactory learning: convergent findings from lesion and brain imaging studies in humans
1Neuropsychology and Cognitive Neuroscience Unit, Montreal Neurological Institute, Montreal, Quebec, Canada H3A 2B4 Correspondence to: Lauren A. Dade, Rotman Research Institute, Baycrest Center for Geriatric Care, 3560 Bathurst Street, Toronto, Ontario, Canada M6A 2E1 E-mail: ldade{at}rotman-baycrest.on.ca
Received November 6, 2000. Revised June 8, 2001. Second revision August 10, 2001. Accepted August 29, 2001. .
 |
Summary |
|---|
 Top Summary IntroductionStudy I. Olfactory memory...Study II. Functional imaging...References |
|---|
The role of temporal lobe structures in olfactory memory was investigated by (i) the examination of odour learning and memory in patients who had undergone resection from a temporal lobe (including primary olfactory regions) for the treatment of intractable epilepsy; and (ii) the examination of brain function during odour memory tasks as assessed via PET imaging of healthy individuals. In order to study different stages of odour memory, recognition of a ‘list’ of odours was tested after a first exposure, again after four exposures and once more after a 24 h delay interval. Patients with resection from a temporal lobe performed significantly less well than control subjects on all trials, and no significant differences were noted as a function of side of resection, indicating that there is not a strong hemispheric superiority for this task. The PET data yielded different levels of activity in piriform cortex (primary olfactory cortex), in relation to the ‘no-odour’ baseline scan, depending on the type of processing: no increase in activity noted during odour encoding, a small increase bilaterally during short-term recognition and a larger increase bilaterally during long-term recognition. These findings, together with findings in animal studies, suggest that piriform cortex may have an active role in odour memory processing, not simply in odour perception. Taken together, the findings from the lesion study and functional brain imaging of healthy subjects suggest that olfactory memory requires input from left and right temporal lobe regions for optimal odour recognition, and that, unlike with verbal or non-verbal visual material, there is not a strong functional lateralization for olfactory memory.
Keywords: olfaction; recognition; PET; short-term memory; long-term memory
Abbreviations: BA= Brodmann area; CAH = corticoamygdalo-hippocampectomy; FER = fourth exposure recognition trial; HERA = hemispheric encoding/retrieval asymmetry; LR = left resection; NC = normal control; PEA = phenylethyl alcohol; rCBF = regional cerebral blood flow; RR = right resection; SAH = selective amygdalo-hippocampectomy; SER = single exposure recognition trial
|
Introduction |
|---|
|
TopSummaryIntroductionStudy I. Olfactory memory...Study II. Functional imaging...References |
|---|
The current understanding of brain regions involved in human olfactory memory is based upon findings from animal studies (for reviews, see Haberly, 1985
; Shipley and Ennis, 1996), human anatomical investigations (Price, 1990), as well as from human brain lesion research (for a review, see West and Doty, 1995). Although currently there are a number of olfactory imaging studies (for reviews, see Zald and Pardo, 2000; Zatorre and Jones-Gotman, 2000), none have looked at encoding, or at short-term versus long-term recognition of learned odours. Our purpose, therefore, was to investigate these aspects of olfactory memory through a convergent approach of examining patients with resection from temporal lobe structures, as well as studying these same aspects in the healthy brain, through the use of PET.
As early as the 1800s, the relationship between human olfactory function and the temporal lobe region was raised in the context of olfactory auras, and olfactory dysfunction was noted in temporal lobe epilepsy patients (Hughlings Jackson and Beevor, 1889; Hughlings Jackson and Stewart, 1899). Anatomical studies in mammals have shown that the olfactory tract projects ipsilaterally to the piriform cortex (the caudolateral aspect of the orbitofrontal cortex, at the frontotemporal junction, extending to the anterior dorsomedial aspect of the temporal lobe), anterior cortical nucleus of the amygdala and the periamygdaloid and entorhinal cortices; all but the orbitofrontal region are temporal lobe structures. From these primary regions, there are important connections to the hippocampus, ventral striatum, thalamus and rostral orbitofrontal (area 11) cortex (for reviews of anatomy, see Eslinger et al., 1982; Price, 1990; Carmichael et al., 1994; Shipley and Ennis, 1996). However, although the olfactory system connects intimately with temporal lobe structures known to be important for memory (Meyer and Yates, 1955; Milner, 1968a, b), it is not clear which specific regions within temporal cortex are most important for olfactory memory. It is also unclear whether there is a hemispheric asymmetry for olfactory memory as has been found to be the case in studies of other types of materials (Meyer and Yates, 1955; Milner, 1958, 1968b; Jones-Gotman et al., 1997b).
Studies have examined olfactory memory in groups of epilepsy patients both before and after surgical intervention in different brain regions, and only subjects with resection within temporal lobe and orbitofrontal regions have shown impairments (for reviews, see West and Doty, 1995; Zatorre and Jones-Gotman, 2000). In studies of temporal lobe epilepsy patients, the results have been unclear as to whether a greater olfactory memory impairment occurs with left or right temporal lobe dysfunction. One study of unoperated epilepsy patients found no impairment on a monorhinal odour recognition paradigm (Eskenazi et al., 1986), while two other studies showed odour memory impairment in patients who had epilepsy arising from either temporal lobe (Martinez et al., 1993; Savic et al., 1997). Consistent with the latter findings, three studies of operated patients showed impairments after resection from either the left or the right temporal lobe (Rausch et al., 1977; Eskenazi et al., 1983, 1986). Interestingly, Eskenazi and colleagues, using a monorhinal recognition paradigm, showed that the recognition deficits in these patients were confined to the nostril ipsilateral to the side of the temporal lobe lesion (Eskenazi et al., 1986). Also, the study by Rausch and colleagues hinted at a greater role of the right temporal lobe in odour memory: although both patient groups were impaired, those with a right resection were significantly more impaired than patients with a left resection (Rausch et al., 1977). In contrast, Henkin and colleagues found that patients with excision from the left temporal lobe performed less well on an odour recognition test than patients with a right resection (Henkin et al., 1977).
Support for greater right hemisphere involvement has come from studies finding odour memory impairment restricted to patients with damage in the right hemisphere (Abraham and Mathai, 1983; Carroll et al., 1993; Jones-Gotman and Zatorre, 1993). Additional support for a right hemisphere advantage in olfactory function arises from other non-memory olfactory findings, such as a right nostril/right hemisphere advantage for detection (Cain and Gent, 1991) and discrimination (Zatorre and Jones-Gotman, 1990, 1991), and a right hemisphere priming effect for odour naming (Zucco and Tressoldi, 1989). The more consistent findings of right hemisphere superiority in these other olfactory processing tasks have given weight to hypotheses about right hemisphere dominance for olfactory memory, but this remains unproven. One tool that can aid in the investigation of hemispheric specialization for olfactory memory is functional brain imaging. By combining this approach with the study of brain-lesioned patients, it is possible to determine the structures that participate in olfactory memory in the healthy brain through imaging, and learn what regions are necessary to perform the task within normal limits through the observation of patients with surgical lesions. Although PET has been used to study other olfactory functions, potential differences in brain activity during encoding and long-term recognition have not been examined.
For these experiments, an olfactory memory test was designed that examined recognition: (i) after a first exposure (a measure of initial encoding); (ii) during recognition after four exposures (to examine short-term recognition and learning); and (iii) during recognition after an extended delay interval (to study long-term recognition and retention). This paradigm was designed for behavioural testing of patients and healthy control subjects, and was adapted for use in a PET research design that would examine healthy brain activity during the cognitive processes of odour encoding and recognition. Areas anticipated to show increased activity during the odour-present tasks (encoding and recognition) compared with the ‘no-odour’ control task included the primary (piriform) and secondary (right orbitofrontal) olfactory cortices. Although early PET findings suggested that activity in primary olfactory cortex (bilateral piriform) represented basic sensory processing, and would therefore be present during all odour-processing tasks (Zatorre et al., 1992), later studies have shown inconsistent levels of piriform activity that cannot be explained simply by differences in research methodology (for discussion, see Zald and Pardo, 2000; Zatorre and Jones-Gotman, 2000). Thus, the presence or absence of activity within piriform cortex would be of interest. Possible differences in activation between left and right prefrontal regions during odour encoding versus recognition were also of interest, as several different hypotheses about their participation in encoding and retrieval have arisen (Tulving et al., 1994; Buckner and Koutstaal, 1998; Kelley et al., 1998; Wagner et al., 1998; Frey and Petrides, 2000). For example, some researchers claim that the left prefrontal cortex is more involved in episodic encoding, while right prefrontal regions are involved in episodic retrieval (hemispheric encoding/retrieval asymmetry, or ‘HERA’, model) (Tulving et al., 1994; Buckner, 1996). Differences in areas of activity between short- and long-term recognition conditions were also of interest, because in other tasks prefrontal regions have been shown to be involved preferentially in short-term as opposed to long-term memory processing (Dade et al., 1997a, 2001). Additionally, areas of temporal lobe activity detected by PET would be compared with the area of brain resection in the patient study, as we predicted that areas showing increased activity during odour-present PET conditions would correspond to regions that, when resected in patients, result in disruption of olfactory memory processing. Portions of these data have been presented previously in abstract form (Dade et al., 1998).
|
Study I. Olfactory memory in patients with resection from a temporal lobe |
|---|
|
TopSummaryIntroductionStudy I. Olfactory memory...Study II. Functional imaging...References |
|---|
Methods
Subjects
Forty patients who had undergone unilateral resection from the left (n = 21) or right (n = 19) temporal lobe for the treatment of intractable epilepsy and 21 healthy normal control subjects (NC) matched approximately for gender, age, education and smoking habits (Table 1) participated in this study. In all cases, the origin of the patient’s epileptic seizures had been localized to a single focus as determined by EEG, MRI and clinical findings. All except four patients were right handed; the left-handed patients had been shown by preoperative intracarotid amobarbital testing (Branch et al., 1964
) to have speech representation in the left cerebral hemisphere. All patients were of normal intelligence, with Full Scale WAIS-R (Wechler Adult Intelligence Scale—Revised) IQ (Wechsler, 1981) ratings above 80. Analysis of variance did not demonstrate any significant differences in age or education level among the three groups, and there was no significant difference in Full Scale IQ ratings between the two patient groups (Table 1). This study was approved by the Research Ethics Committee of the Montreal Neurological Institute and Hospital. All subjects gave informed written consent according to the Declaration of Helsinki.
|
Patients had undergone one of two different surgical approaches, both involving removal of mesial temporal lobe brain structures. The majority of patients [left resection (LR) = 15; right resection (RR) = 13] underwent a corticoamygdalo-hippocampectomy (CAH) or anterior temporal lobe resection. The surgical CAH approach consists of a neocortical resection (averaging between 4.5 and 5.0 cm along the sylvian fissure, extending to the level of the precentral sulcus), partial to complete resection of amygdala and varying extents of excision from hippocampus and parahippocampal gyrus (ranging from 0.5 to
3 cm) (Trop et al., 1997). The remaining 12 patients (LR = 6; RR = 6) underwent a selective amygdalo-hippocampectomy (SAH). This approach involves making a small 2–3 cm excision through the second temporal gyrus, just below the superior temporal sulcus. The sulcus is followed down to the floor of the lateral ventricle, where resection of the amygdala (partial to complete) and hippocampus (0.5–2.5 cm) takes place, leaving the neocortex relatively intact (Trop et al., 1997).
Determination of the exact extent of an individual patient’s resection is difficult, and intraoperative reports often overestimate the area removed (Awad et al., 1989). Therefore, for the purpose of verification of the area of surgical excision, measurements of the resection were made by examining postoperative T1-weighted MRI scans obtained using 1 mm slice acquisition. Of the patients participating in the study, 21 had the appropriate scan available for measurement (LR = 10; RR = 11). Each scan was transformed into Talairach space (Talairach and Tournoux, 1988) using an automated algorithm and the MNI (Montreal Neurological Institute) average template of 305 healthy subjects as the target space (Collins et al., 1994). Measurements were taken by progressing coronally along the y-axis through each millimetre slice image of the entire brain scan. The extent of remaining tissue of the hippocampus, and entorhinal and parahippocampal cortices was calculated as the length of the structure from the most rostral point of intact tissue to the most caudal point (Crane, 1999) (see Table 2). A gross approximation of the extent of excision from the amygdala was made by a visual comparison with the unoperated hemisphere and estimating the percentage of tissue that was remaining. In the CAH patients, the extent of removal along the superior and inferior temporal gyri was estimated as the difference between the y coordinate of the intact temporal pole and the y coordinate of the posterior-most point of the lesion. The extent of excision averaged 22 mm along the superior temporal gyrus (range 5–50 mm) and 43 mm along the inferior temporal gyrus (range 29–70 mm).
|
Precise anatomical landmarks for the anterior and posterior borders of the piriform cortex, which can be determined from an MRI scan, have not been defined. However, we wanted to be able to quantify the extent of resection within this region in the patients for whom we had MRI scans. Therefore, we chose a novel approach of estimating the extent of piriform excision by identifying it via the existing functional imaging information about the piriform area activations that occur during smelling. Thus, the piriform region was identified by utilizing the averaged PET coordinates for piriform activity obtained from the current PET study and previous studies (Zatorre et al., 1992
, 2000; Small et al., 1997; Sobel et al., 1998; for discussion, see Zatorre and Jones-Gotman, 2000). The limits of left and right piriform regions were defined as 2 SD above and below the average left (x = –21, y = 5, z = –19; 1 SD: x = 2, y = 4, z = 3) and right (x = 21, y = 4, z = –14; 1 SD: x = 3, y = 6, z = 1) Talairach coordinates (Talairach and Tournoux, 1988) of PET piriform activity. Our definition of piriform cortex therefore takes into account the spatial blurring introduced by PET, and extends beyond anatomically defined piriform cortex (which is very small) to include surrounding tissue that responds functionally to the presence of odours. Using our patient MRIs transformed (Collins et al., 1994) into Talairach space (Talairach and Tournoux, 1988), we were able to examine each coronal MRI image at 1 mm intervals between the 2 SDs above and below the y coordinate for the left and right piriform cortices. At each 1 mm coronal slice image, the medial, lateral, superior and inferior margins of the piriform area (as determined by the 2 SD above and below the averaged x and z coordinates) were examined to determine if tissue was present or excised. The numbers of coronal slices where piriform was intact were then added together as total intact millimetres and divided by the total extent of the piriform region in the coronal plane as identified by the PET measurements (left piriform total coronal extent = 18 mm; right piriform total coronal extent = 24 mm). The percentage of intact piriform tissue within the specified range is reported in Table 2. All patients had removal from the mesial temporal lobe structures (amygdala, hippocampus, entorhinal and parahippocampal cortices) and resection from one or more of the olfactory regions (piriform cortex, amygdala). As both surgical approaches involved the olfactory regions of interest, and due to the small number of SAH subjects (which would limit the statistical power of an analysis), SAH and CAH patients were combined, and patients were assigned to two groups based on the side of resection: LR or RR.
Stimulus materials
Odours were provided by the Givaudan-Roure Corporation (Teaneck, NJ, USA). All odours were presented dirhinally via puffs of scented air, presented 1 inch below the nose, from opaque squeeze bottles in a format modified from Cain et al. (1992). Forty-eight odours arising from 12 categories (four from each category) were used (Table 3). In order to reduce the effectiveness of verbal labels, the distractor odours to be used in recognition testing were chosen from the same categories as the to-be-learned (target) items. For example, if a subject labelled a target odour as ‘fruity’, this label by itself would be ineffective, because a new ‘fruity’ distractor odour would also occur during recognition. In order to ensure that subjects could discriminate among odours arising from the same category, and among odours from a similar category, pilot testing for odour discriminability was carried out. The odours were divided into six comparison sets, each consisting of two similar categories: fruity/citrus, woody/balsam wood, heavy floral/light floral, minty/grassy, spice/anise and animal-like/unpleasant. Six groups of 20 healthy control subjects (60 men and 60 women in total) participated in the discrimination tests. A three-choice oddball paradigm was used, with dirhinal odour presentation; subjects stated which odour of the three presented odours was the different one. Based on the results of the pilot study, the most highly discriminable odour from same-category odours and close-category odours was selected to be a target odour in the memory experiment. This resulted in 12 target odours for the memory test, one from each of the odour categories. The mean percentage correct discrimination of all the target odours from other odours in the same category and from odours in a closely related category was 85% (range 70–95%; median 85%; mode 90%). The three distractor odours from each category were distributed randomly across three recognition trial odour sets (A1, A2 and A3). Three testing forms were created such that, across forms, each recognition set was used at each of the different testing intervals (first recognition, second recognition and 24 h delay). Use of the different test forms was counterbalanced across subjects.
|
Procedure
All subjects were screened for normal odour detection thresholds using a modified monorhinal two alternative forced-choice detection of phenylethyl alcohol (PEA) versus water. Subjects passed this screening if they were able to detect the PEA correctly on four consecutive trials.
For the odour memory test, during the encoding phase, the 12 target odours were presented serially with a 20 s interstimulus interval. Subjects were told to smell the odour and try to remember it. They were advised that they would never be asked to name an odour. Three minutes after the last target was presented, subjects underwent the single exposure recognition trial (SER) with presentation of 24 odourants: the 12 target odours interspersed amongst 12 distractor odours. Subjects gave a ‘yes/no’ recognition response for each item. This was followed by three subsequent presentations of the target odours in three pseudorandom orders, for learning, and then the fourth exposure recognition trial (FER) using 12 new distractor odours. Subjects then returned the next day for a 24 h delayed recognition test, with the 12 target odours interspersed amongst the third set of distractor odours. Scores were obtained by adding correct hits (maximum = 12) and correct rejections (maximum = 12) for a total possible score of 24 on each test.
Results
The average scores and standard deviations on the SER, FER and 24 h delay were: NCs = 18.7 (2.5), 21.1 (1.8), 20.1 (2.5); LR = 16.1 (2.7), 18.5 (3.1), 16.7 (3.5); RR = 15.6 (2.5), 18.7 (2.5), 16.8 (2.8). A three-way repeated measures analysis with two between-group factors [group (LR, RR, NC) and gender] and one within-subjects factor (test: SER, FER, 24 h delay) revealed a significant effect of group [F(2,55) = 10.2, P < 0.0002] and test [F(2,110) = 22.4, P < 0.0001]. There was no effect of gender and there were no significant interactions. Post hoc Tukey HSD tests for unequal numbers revealed that the RR and LR groups were not significantly different from each other. However, both patient groups performed significantly less well than the normal control group (P < 0.001). Neither the LR nor the RR group performed significantly better than chance on the SER trial [LR: t(18) = 1.53, P > 0.05; RR: t(18) = 1.46, P > 0.05]. A post hoc Tukey test of the main effect of trial showed that performance on the FER trial was significantly better than performance on the SER trial (P < 0.001) and performance after a 24 h delay (P < 0.001). Recognition memory after a 24 h delay was significantly better than recognition on the SER trial (P < 0.006) (Fig. 1).
|
Discussion
In accordance with a number of the previously mentioned olfactory memory studies (Rausch et al., 1977
; Eskenazi et al., 1986; Martinez et al., 1993; Savic et al., 1997), no significant differences were noted between patient groups as a function of side of temporal lobe resection. In fact, the similarity of the performance across trials for these two groups was striking, and both were impaired compared with healthy control subjects. There was no significant interaction of group by trials, indicating that the pattern of performance across the three trials was the same for the three groups. Performance of the LR and RR groups on the SER was at 67% correct, while the NC group performed at 79% correct, with little information lost from the FER to 24 h delayed recall (NC group lost 4%, left and right groups lost 8%). The results for the long-term retention are in keeping with the findings of Engen and Ross (Engen and Ross, 1973) and Lawless and Cain (Lawless and Cain, 1975), which showed imperfect initial recall after a single exposure (70–85% recognition immediately after all odours were presented) with a forgetting curve that did not exceed a 10% loss of information even after a 30 day delay (for a review see Schab and Crowder, 1995).
Initial odour encoding in the patient groups was poor and, although they were able to learn with additional exposures, they were not able to recover from this initial deficit. One possible explanation for their poor performance is that they were impaired in discriminating among the different odours presented. Discrimination deficits have been noted in patients with temporal lobe lesions; however, these deficits are relatively mild and are restricted to the nostril ipsilateral to the side of lesion (Zatorre and Jones-Gotman, 1991). Therefore, employing a dirhinal paradigm combined with the use of odours that have been tested for their high discriminability should lead to adequate discrimination performance for this type of recognition test. Indeed, if these patients did have some difficulties in discrimination, it did not impair their ability to learn the odours, as their learning curve closely mirrored the performance of the healthy control subjects.
The issue of verbal labelling was also addressed in the current paradigm. The question regarding the use of verbal labelling of odours is particularly important in lesion studies, as patients with a right hemisphere lesion may be able to ameliorate their performance by relying on their intact verbal skills. Hence, if right temporal lobe epilepsy patients are able to apply a verbal strategy effectively, this may contribute to the inconclusive findings across studies with regard to right hemisphere specialization. Although it is difficult to give precise labels for odours (Cain, 1979; Carroll et al., 1993; Lehrner et al., 1999a), a general category label may suffice as an additional memory cue. However, in the current study, the utility of applying general category labels was rendered much less effective by requiring subjects to recognize specific odours they had smelled before (targets) from among other odours that derived from the same general categories (distractors). Therefore, if the right temporal lobe is more involved in odour memory processing than the left, and if RR patients cannot compensate through labelling for losing this advantage (as in the present paradigm), one should see a greater disparity between the performance of LR and RR groups. However, this was not the case; both patient groups showed a clear and equal deficit, suggesting that both temporal lobes, perhaps specifically the piriform regions, play an important role in odour memory.
|
Study II. Functional imaging of healthy volunteers |
|---|
|
TopSummaryIntroductionStudy I. Olfactory memory...Study II. Functional imaging...References |
|---|
To gather converging evidence regarding the role of the piriform cortex in olfactory memory, and to gain additional understanding of other brain regions involved in olfactory memory, we used the functional imaging technique of PET to study odour memory processing in the healthy brain. By applying this complementary approach of studying brain function in subjects with resection from a temporal lobe, and examining brain function in healthy subjects through imaging, a more comprehensive view can be taken that circumvents some of the limitations of either approach. For example, in patient studies, the area of brain lesion or resection typically involves a number of different anatomical structures (Jones-Gotman et al., 1997
a; Trop et al., 1997), and the effects of disconnection between brain areas are unclear. Thus, although it is possible to learn what broad brain regions are necessary to perform a particular task, this approach does not allow examination of the function of specific brain structures within and outside that brain region. On the other hand, brain imaging can give us more specific information about the different structures that may be involved in olfactory recognition, as well as insight into differences in functional participation of a particular region across a range of cognitive processes (e.g. encoding, short-term and long-term recognition). By melding these two approaches, it is possible to investigate the performance of brain regions during different odour memory processes, and to learn what regions need to be intact in order to recognize odours without impairment.
Subjects were scanned while performing a task similar to that of Study I. Three odour-processing scans were performed: odour encoding, short-term recognition and long-term recognition. Subjects were also scanned during a no-odour baseline condition, which acted as a control for motor movements and air sensations against the face. By subtracting the regions of brain activity occurring during this baseline condition from each of the three odour-processing scans, we were able to examine the brain regions involved specifically in odour encoding and recognition. In order to compare differences in activity that may occur between short- and long-term recognition, the two PET regional cerebral blood flow (rCBF) recognition images were subtracted from each other. Predicted areas of activity during odour processing included: bilateral piriform and right orbitofrontal cortices (related to odour processing), temporal lobe activity (related to memory components) and differential activity in left and right prefrontal areas (related to encoding and/or recognition processes) (Tulving et al., 1994; Buckner, 1996; Wagner et al., 1998).
Methods
Subjects
Twelve healthy right-handed volunteers (six men, six women) participated in this study (mean age 24.8 years; range 20–30 years). None had a previous history of neurological or psychiatric disorders. All subjects were non-smokers with no history of nasal injury. This study was approved by the Research Ethics Committee of the Montreal Neurological Institute and Hospital. Subjects were paid for their participation and gave informed written consent according to the Declaration of Helsinki.
Stimulus materials
Thirty-six of the 48 odours from Study I were used in this experiment: the 12 target odours and 24 of the distractor odours.
To equate for level of difficulty, the 12 target and 24 distractor odours were assigned to two target odour sets and four distractor odour sets such that mean discriminability scores were equivalent across recognition tests. Odours were presented dirhinally in puffs of air via opaque squeeze bottles. The baseline condition consisted of puffs of air without any odour added (Fig. 2). Odours were presented in the same way during the training session and during the PET scan: each odour was presented for 4 s followed by a 6 s interstimulus interval.
|
Procedure
Training session
Subjects participated in a training session 4 days prior to the PET study, in order to learn the odours for the long-term odour recognition scan. Subjects were instructed to smell the six target odours and try to remember them. They were informed that they would never be asked to name the odours, and that they would just have to recognize the odours presented among others in a larger set. Subjects were also told to sniff with the same depth and frequency of inhalation throughout all tasks. Training consisted of an initial presentation of the six long-term target odours, a first recognition test (Recognition 1), then three serial presentations of the target odours (in pseudorandom order) as learning trials. A second recognition test (Recognition 2) followed the last presentation of targets (Fig. ). Recognition trials involved the pseudorandom presentation of the six target and six distractor odours. Yes/no recognition responses were given via key-press following each stimulus presentation.
PET
Subjects returned 4 days following the training session and were scanned during four conditions: (i) no-odour sensorimotor control task; (ii) long-term odour recognition; (iii) encoding of new odours; and (iv) short-term odour recognition of new odours. All scans occurred in the same order and included six odour presentations, at the rate of one odour each 10 s, over the 60 s scanning period. Each recognition task consisted of 12 stimulus presentations: six presentations occurred during scanning, with four or five of the target stimuli presented (and one or two distractor stimuli) during the PET data acquisition window to ensure that the results would reflect potential odour recognition. Yes/no responses were given via key-press. Subjects were reminded to sniff in the same way throughout all scans including the baseline control task and to keep their eyes closed during each condition. To avoid subjects engaging in an odour search task (and possibly increasing activation in odour-processing regions) during the no-odour condition, subjects were informed that they would feel the puffs of air, but that no odour would be present. In this condition, subjects made random key-press responses following each air-puff presentation to control for motor function. The six odours presented during the encoding scan were used as the target odours for the short-term recognition scan. Between the encoding and short-term recognition scans, the short-term target odours were presented serially three additional times, to approximate the number of odour exposures subjects had experienced with the long-term target odours prior to recognition (Fig. ).
Data acquisition and analysis
PET scans were obtained using a CTI/Siemens HR+ 63 slice tomograph with an intrinsic resolution of 4.2 x 4.2 x 4.2 mm. rCBF was measured during a 60 s scan using the [15O]water bolus method (Raichle et al., 1983). MRI scans were obtained using either a Siemens Vision (1.5 T) or a Philips ACS III (1.5 T) scanner. Both produced a high-resolution three-dimensional whole brain T1-weighted scan (140–160 sagittal slices of 1 mm). PET and MRI scans were co-registered and resampled into standardized stereotaxic space as defined by Talairach and Tournoux (Talairach and Tournoux, 1988) (for methodology see Evans et al., 1992). Images were reconstructed using a 14 mm Hanning filter and averaged across subjects for each scanning condition. Differences in rCBF were examined by carrying out a paired image subtraction of scans of interest. The significance of focal CBF changes was evaluated using a method based on three-dimensional Gaussian random field theory (Worsley et al., 1992). A threshold for significant t-statistic peaks was set at t 
±3.53 (P 
0.0004) for grey matter volume of 500 cm3 and 182 resolution elements (14 x 14 x 14 mm), yielding an expected false-positive rate of 0.58 (Worsley et al., 1992). For the principal olfactory regions (piriform cortex and right orbitofrontal cortex, insula), where activity was predicted based on previous findings (Zatorre et al., 1992), the threshold was lowered to t = 3.0.
Results
Behavioural results
Subjects performed well on all recognition tests (mean percentage correct during training: Recognition 1 = 70%, Recognition 2 = 84%; mean percentage correct during PET: long-term delayed recognition = 76%, short-term recognition = 85%). A one-way repeated measures analysis of variance for recognition tests 1 and 2 and 4-day delayed recognition showed a significant difference in the percentage correct [F(2,22) = 6.53, P < 0.01]. Post hoc Tukey tests revealed that performance on the FER, after learning, was significantly better than on the SER (P < 0.01). Performance on the long-term recognition trial was not significantly different from trials 1 or 2. A paired t-test comparing performance between the short- and long-term recognition PET tasks revealed marginally better performance on the short-term recognition task [t(11) = 1.76, P = 0.05; one-tailed].
PET results
Long-term recognition minus baseline
Significant activity was noted in primary (bilateral piriform cortices) and secondary (right orbitofrontal) olfactory cortices, as expected. These regions were close to the areas previously reported (Zatorre et al., 1992, 2000; Small et al., 1997; Zald and Pardo, 1997; Sobel et al., 1998). Significant activity was also noted bilaterally in insular cortex.
Significant prefrontal activity was detected in bilateral cingulate regions (BA 32), an area detected in other encoding and retrieval studies (Buckner, 1996), and bilateral medial frontal [Brodmann area (BA) 8] cortex. Outside these areas of interest, activity was noted in right caudate, left precentral gyrus and right inferior parietal lobe (Table 4 and Fig. 3).
|
|
Odour encoding minus baseline
This subtraction was expected to reveal areas important for odour processing (piriform and orbitofrontal cortices) and perhaps greater left prefrontal activation activity related to encoding processes as proposed in the HERA model (Tulving et al., 1994
; Cabeza et al., 1997). No significant activation was noted in primary (piriform) olfactory cortex. A region of subthreshold activity was noted near the right piriform (t = 2.89; x = 21, y = 6, z = –11). However, its anomalous location, >3 SD dorsal to the average z coordinate for our defined piriform region, and its low t value, make it difficult to interpret this finding. No significant activity was noted in the secondary (right orbitofrontal cortex) olfactory region. Significant activations were found only in bilateral superior (t = 4.2; BA 8; x = 3, y = 27, z = 59) and medial (t = 4.0; BA 6; x = 1, y = –14, z = 57) frontal cortices, as well as in the left precentral gyrus (t = 3.9; BA 4; x = –11, y = –32, z = 66).
Short-term recognition minus baseline
Significantly increased activity was present in bilateral piriform, right orbitofrontal (BA 11) and insular cortices during short-term recognition. Activity was also noted in dorsolateral prefrontal cortex, a region known to play a role in working memory (Petrides, 1994a, b; Goldman-Rakic, 1995). Additional activity occurred within bilateral pre- and postcentral gyri and within the left lingual gyrus (Table 4 and Fig. ).
Long-term recognition minus short-term recognition
In the regions of interest, significantly greater activity was noted in right piriform cortex, left insula and left mid-dorsolateral prefrontal cortex during long-term recognition. During short-term recognition, significantly greater activity was noted in left and right mid-dorsolateral frontal cortex, and the left frontal polar region. The differences in piriform and right mid-dorsolateral frontal activity between the two conditions can be noted in Fig. , where the two tasks are compared with the same baseline condition. Additional regions of difference were found in parietal lobe and motor/sensory regions of the frontal lobe (see Table 5).
|
Relationship between PET results and extent of excision in patients
All patients had excision within at least one, and usually more than one, primary olfactory region (piriform cortex, amygdala, periamygdaloid and entorhinal cortices). Of patients who had MRI measurements, all but one had excision from the piriform cortex and all had excision from amygdala. To examine the relationship between the region of piriform activity detected during functional imaging and the region of temporal lobe resection in the patients, the long-term recognition PET data were co-registered in Talairach space (Talairach and Tournoux, 1988
; Collins et al., 1994) with the structural MRI of a patient with a standard CAH resection. The area of odour recognition activity obtained in the healthy subjects overlaps, in part, with the excised temporal lobe region, further indicating that important odour-processing areas are invaded in the patient population (Fig. 4).
|
General discussion
The most interesting PET findings were the differences in piriform activation across the three odour-processing conditions. The lack of piriform activity during encoding was notable, particularly considering that in the short-term recognition condition, where essentially the same odours were presented, piriform activity was observed in comparison with the same baseline condition. This modulation in piriform activity has now been seen across different studies: the first 15O-PET study of human olfaction (Zatorre et al., 1992
) showed strong bilateral piriform activity during smelling. However, a lack of increased piriform activity during smelling has been noted in other studies (Dade et al., 1997b; Yousem et al., 1997; Sobel et al., 1998; Zatorre et al., 2000), and one study found only subthreshold changes that did not reach pre-set levels of significance (Zald and Pardo, 1997). Although increased piriform activity, related to smelling, was not found consistently in early fMRI (functional MRI) studies (Sobel et al., 1998), this issue was resolved (Sobel et al., 2000b) by using a statistical approach that took into account the early transient increase in signal amplitude and response habituation that occurs in piriform regions when the same odour is presented repeatedly (Wilson, 1998a). Using this approach, Sobel et al. (2000b) were able consistently to observe odour-induced activity, in addition to sniffing activity (Sobel et al., 1998), in piriform and ventral temporal lobe regions.
The temporal aspects revealed by fMRI analysis do not entirely explain the differences in detection of piriform activity in the current study, as PET methodology is less sensitive to transient changes in signal amplitude. PET acquires data cumulatively over a 1 min scanning interval, rather than based on responses matched to the time course of stimulus presentation. Also, the factor of response habituation, due to subjects experiencing the same odour repeatedly during the same block (Sobel et al., 1998, 2000b), would not apply to the current study. In one fMRI study, habituation was seen to occur after five presentations of the same odour (Sobel et al., 2000b), a result similar to that found in electrophysiological studies of rats (Wilson, 1998a, b). Findings from rat studies have also indicated that habituation in the piriform is odour specific (Wilson, 1998a). Therefore, it is probable that our paradigm, which consisted of the presentation of a series of different odours during each scan, would not cause habituation in piriform cortex, and our analysis would also not be susceptible to this factor.
The possibility of differences in sniff rate across the different conditions is also an important issue, as piriform activity is influenced by this behaviour (Sobel et al., 1998, 2000b). Although we were unable to measure sniff rate directly, subjects were told always to sniff in the same way (depth and rate of inhalation) for each condition (both in training and prior to each scan), regardless of the presence or absence of an odour. Small differences (of the order of tenths of seconds) in sniff duration have been noted between presentation of low and high concentration of odours (Laing, 1983; Sobel et al., 2000a); however, all the odours used in the present experiment were at suprathreshold detection levels, and there was no bias for lower concentration odours in any one scanning condition (particularly between the encoding and short-term conditions where five of the six odours were the same). Given that subjects were sniffing during our baseline condition, that the pattern of stimulus presentation was the same across all conditions and that subjects were told to sniff in the same way during all scans, it does not appear that changes in rates of sniffing can explain our findings adequately.
Piriform cortex: a memory processor?
An intriguing finding is that in the same subjects, comparing with the same baseline scan, we observed different levels of piriform activity in relation to the cognitive conditions of the task. Notably, the activation in piriform cortex appears to follow along a continuum between the encoding condition, with no significant activity present, to the short-term recognition condition, with weak bilateral activity, to the long-term recognition condition, which shows strong bilateral piriform activity. Hence we see an increase in activity as a function of the recognition components of the task and perhaps in relation to odour familiarity. These findings are in agreement with the proposal of Haberly (Haberly, 1985; Haberly and Bower, 1989) and others (Bower, 1991; Hasselmo and Barkai, 1995), which suggests that the primary olfactory cortex serves as a type of associative memory system, which allows for the association of odour stimuli with memory traces of previously experienced scents.
Initially, these inferences may not appear to be in accordance with the study by Zatorre et al. (1992), in which subjects were scanned during passive smelling of different odours, and strong bilateral piriform activity was noted. Nevertheless, there do appear to be factors that would contribute to a long-term memory processing aspect. Although the odours used in the study of Zatorre and colleagues were difficult to name, most were common household products, and all were selected to be moderately to highly familiar (Zatorre et al., 1992). Also, subjects were familiarized with the odours prior to the scanning condition. Perhaps due to these two manipulations, when subjects smelled the familiar and recently smelled odours, activation of short- and long-term memory networks occurred within piriform regions.
Several findings lend support to the theory that piriform cortex is involved in learning and memory. First, synaptic long-term potentiation, which is known to be important in ‘hippocampal’ memory (for a review, see Bliss and Collingridge, 1993), has also been shown to occur in rat piriform cortex in vitro (Jung et al., 1990; Kanter and Haberly, 1990; Jung and Larson, 1994) and in vivo at the conclusion of learning (Roman et al., 1993; Litaudon et al., 1997). Interestingly, simple exposure to olfactory stimulation does not appear to be sufficient to effect change in piriform activity. For example, rat studies (Roman et al., 1993; Litaudon et al., 1997) showed that alterations in piriform activity did not occur until the significance of the olfactory stimulation had been learned. Other evidence relating piriform cortex to memory function comes from the evaluation of single-neurone activity: different cells were detected which fired in association with either the physical characteristics of the odour, the odour’s reward value or its significance in relation to past events (Schoenbaum and Eichenbaum, 1995). Consequently, this evidence supports the concept that piriform cortex can serve a mnemonic function, as well as a perceptual one.
The finding that piriform activity changes in rats after odour learning (Roman et al., 1993; Litaudon et al., 1997) may also explain the greater piriform activation during recognition than in encoding. Perhaps with multiple odour exposures, or ‘learning’, larger responding networks were created, thus increasing cellular activity. Activity in piriform did not appear to increase with improved performance, as subjects performed better during short-term recognition than long-term recognition. However, the greater piriform activity during long-term recognition could reflect increases related to memory consolidation processes. Memory consolidation, or the formation of long-term memories, is dependent on the passage of time and is theorized to be linked to the cellular mechanisms underlying long-term potentiation (for a review, see McGaugh, 2000), which have been shown to occur in piriform cortices. The importance of piriform regions to memory performance was also supported by the patient findings in Study I, as the resections in the patient groups encroached on piriform cortex, and the patients were equally impaired across all three recognition tests. Further, the combined findings from the patient study and the PET results of bilateral piriform participation during long-term recognition support the idea of an interactive role between left and right piriform regions in order to sustain normal odour recognition.
Odour recognition: a dual hemisphere task
Findings from patient studies showing deficits restricted to the nostril ipsilateral to the lesion (Eskenazi et al., 1986; Jones-Gotman et al., 1997a) and from studies with commissurotomized subjects (Gordon and Sperry, 1969) suggest that each hemisphere is able independently to perceive and recognize familiar odours. However, this does not preclude the existence of a more integrated and higher order cognitive system. In a study of odour recognition in healthy subjects, no differences were shown between the left and right nostrils, but dirhinal scores were significantly higher than monorhinal scores (Bromley and Doty, 1995). The superiority of dirhinal stimulation during an odour recognition task is consistent with findings for odour identification that showed mild identification deficits after resection from either the left or right temporal lobe (Jones-Gotman and Zatorre, 1988; Jones-Gotman et al., 1997a).
Therefore, although evidence suggests that each hemisphere can function independently (Gordon and Sperry, 1969), it would appear that it is the interaction of the two that allows for optimal performance for more complex olfactory processing. One factor that may result in the improvement of function with bilateral participation is increased perceptual acuity, as described by Sobel and colleagues, who showed that each nostril conveys different details about olfactory stimuli to the brain (Sobel et al., 1999). Therefore, it could be the case that slightly different olfactory percepts are encoded within each hemisphere, and it is some complex combination of information within the two piriform cortices that allows for more precise olfactory recognition.
Indeed, on a monorhinal odour recognition task, compared with single odour presentation, Savic and colleagues noted significantly greater right but not left piriform activity, following a 1 h delay (Savic et al., 2000). However, it is difficult to determine if these results are due to a less complex olfactory percept due to the monorhinal presentation, whether the right piriform region is truly more involved in odour recognition or whether this is related to the intermediate time delay, which may not have been sufficient to have allowed the consolidation processes (McGaugh, 2000) that would engage both piriform regions.
Differences in short- and long-term recognition processing
In contrast to the strong piriform activity during long-term recognition, the short-term recognition condition showed greater activity in right mid-dorsolateral prefrontal cortex and parietal lobe, areas known to be important for working memory processing in the visual modality (Petrides, 1994a, b; Goldman-Rakic, 1995). These same regions have been shown to be active during functional imaging of an olfactory working memory task (Dade et al., 2001). Hence these results demonstrate a differentiation between short- and long-term memory processing, and suggest that short-term recognition tasks and working memory tasks (requiring on-line information monitoring and manipulation) engage some of the same anatomical regions.
Insular participation
Insular activity was detected unilaterally on the right during short-term recognition and bilaterally in the long-term recognition condition. The insula is known to receive axonal projections from primary and secondary olfactory regions (Price, 1990; Price et al., 1991), and similar insular and peri-insular activity has been detected during other olfactory functional imaging studies (Zatorre et al., 1992; Zald and Pardo, 1997; Fulbright et al., 1998; Sobel et al., 1998; Savic et al., 2000). The insula is thought to be important for gustation (Shipley and Ennis, 1996; Small et al., 1997, 1999), and activation during this olfactory task may be reflective of the olfactory role in flavour perception.
Orbitofrontal regions
A very consistent finding in the olfactory imaging literature (for reviews, see Zald and Pardo, 2000; Zatorre and Jones-Gotman, 2000) and in the recognition conditions of the current study is the unilateral right orbitofrontal activity. The orbitofrontal region observed in the short- and long-term recognition conditions corresponds most closely to the lateral posterior orbital cortex identified in the monkey (Tanabe et al., 1975b). Cells in this region were found to respond more selectively to different odours than cells in either the piriform or amygdala regions (Tanabe et al., 1975a), and findings from this animal study, and from studies of patients with orbitofrontal lesions (Potter and Butters, 1980; Zatorre and Jones-Gotman, 1991), suggest that the orbitofrontal region plays an important role in odour discrimination.
However, it is curious that, as with the piriform region, no orbitofrontal activity was noted during the encoding condition. The presence of orbitofrontal activity during recognition, and the lack of significant activity during encoding, may be related to the greater odour discrimination demands required during recognition (discriminating between serially presented target and distractor odours), which are not requisite during initial encoding. Also the episodic memory component of recognition introduces an additional factor. For example, Royet and colleagues found greater right orbitofrontal activity when subjects had to make judgements about the familiarity of odours versus judgements of edibility (which would not necessarily involve episodic memory), suggesting a possible long-term odour recognition component to the functioning of the right orbitofrontal region (Royet et al., 1999). Further investigation is necessary, using more closely controlled comparisons that require equivalent levels of discrimination in order to tease apart activity due to increased discrimination demands (e.g. choosing between closely matched odours) versus increased demands of retrieval (e.g. long delay intervals).
Prefrontal cortical involvement
Activity in prefrontal cortex also needs to be considered in terms of processing demands. The HERA model broadly suggests that left prefrontal cortex is involved differentially in encoding, while right prefrontal cortex is involved differentially in retrieval (Tulving et al., 1994). A limitation of this model is that it does not describe particular anatomical regions within prefrontal cortex that perform these functions. In a more detailed analysis, specific prefrontal regions involved in encoding (left inferior frontal BA 44 and BA 45 during verbal encoding) and retrieval (right anterior frontal BA 10) were identified (Buckner, 1996). In the current study, the most prominent source of prefrontal activity is within the right orbitofrontal region during recognition. Thus, it would seem that orbitofrontal activity (BA 11) is not usually present in studies of information retrieval and, in the current paradigm, its function is directly related to the olfactory recognition demands of the task. Findings of significant increases in blood flow in right BA 10 and right BA 9/46 during the short-term odour recognition condition do fit with the above model, and also with activity detected during an olfactory working memory task (Dade et al., 1997b). However, these same regions are not active during long-term odour recognition, which suggests that working memory areas can be engaged during short-term memory processing, but that these regions are not relied on during long-term memory retrieval.
Conclusion
When using an odour recognition paradigm that controlled for the confound of verbal labelling, no hemispheric superiority was found among patients with resection from the left or right temporal lobe regions. These findings, combined with imaging data from healthy subjects, support a dual hemisphere role in odour memory processing. In concert with findings from animal studies (Roman et al., 1993; Litaudon et al., 1997), our findings suggest that the role of the piriform cortex extends beyond ‘simple’ sensory processing and reaches into the realm of memory function.
|
Acknowledgements |
|---|
We wish to thank the subjects who participated in this experiment, the Givaudan-Roure corporation for the donation of the olfactory stimuli, and Sylvain Milot and the staff of the McConnell Brain Imaging and Medical Cyclotron Units for their assistance. We also wish to thank J. Crane for her ongoing input and guidance with the measurements of the patient’s surgical resections, and D. Klein and D. McMackin for reviewing earlier versions of this manuscript. This work was supported by Grant MT144991 from the Canadian Institutes of Health Research to M.J.-G. and R.J.Z., and by the McDonnell-Pew Cognitive Neuroscience Program.
|
References |
|---|
|
TopSummaryIntroductionStudy I. Olfactory memory...Study II. Functional imaging...References |
|---|
Abraham A, Mathai KV. The effect of right temporal lobe lesions on matching of smells. Neuropsychologia 1983; 21: 277–81.[ISI][Medline]
Awad IA, Katz A, Hahn JF, Kong AK, Ahl J, Lüders H. Extent of resection in temporal lobectomy for epilepsy. I. Interobserver analysis and correlation with seizure outcome. Epilepsia 1989; 30: 756–62.[ISI][Medline]
Bliss TV, Collingridge GL. A synaptic model of memory: long-term potentiation in the hippocampus. [Review]. Nature 1993; 361: 31–9.[Medline]
Bower JM. Piriform cortex and olfactory object recognition. In: Davis JL, Eichenbaum H, editors. Olfaction: a model system for computational neuroscience. Cambridge (MA): MIT Press; 1991. p. 265–86.
Branch C, Milner B, Rasmussen T. Intracarotid sodium amytal for the lateralization of cerebral speech dominance: observations in 123 patients. J Neurosurg 1964; 21: 399–405.[ISI][Medline]
Bromley SM, Doty RL. Odor recognition memory is better under bilateral than unilateral test conditions. Cortex 1995; 31: 25–40.[ISI][Medline]
Buckner RL. Beyond HERA: contributions of specific prefrontal brain areas to long-term memory retrieval. Psychonom Bull Rev 1996; 3: 149–58.
Buckner RL, Koutstaal W. Functional neuroimaging studies of encoding, priming, and explicit memory retrieval. [Review]. Proc Natl Acad Sci USA 1998; 95: 891–8.
Cabeza R, Grady CL, Nyberg L, McIntosh AR, Tulving E, Kapur S, et al. Age-related differences in neural activity during memory encoding and retrieval: a positron emission tomography study. J Neurosci 1997; 17: 391–400.
Cain WS. To know with the nose: keys to odor identification. Science 1979; 203: 467–70.
Cain WS, Gent JF. Olfactory sensitivity: reliability, generality, and association with ageing. J Exp Psychol Hum Percept Perform 1991; 17: 382–91.[ISI][Medline]
Cain WS, Cometto-Muñiz JE, de Wijk RA. Techniques in the quantitative study of human olfaction. In: Serby MJ, Chobor KL, editor