Locational firing o f pvramidal cells (place cells and place fields)
The most basic observation is that complex spike cells, which have been shown to be the pyramidal cells in the CA3 and CA l fields, fire at relatively high rates in a
specific portion o f an environment, and relatively low rates outside this portion o f the environment. The specific portion o f the environment where a given cell fires at relatively high rates can be termed the place field o f the cell, which is itself termed a place cell (O ’Keefe, 1976).
Basic Place field characteristics
Spatial measurement paradigms and statistics have not developed sufficiently to describe place fields quantitatively to the satisfaction o f all investigators. Some general properties however can be noted. It is clear that the distribution o f firing within place fields approaches a 2-dimensional gaussian distribution (Breese et al, 89; M uller et al, 87; Muller et al, 1994; O ’Keefe and Burgess, 1996), with peak firing describing a centre in the field, and lower firing describing its edges. Figure 4 o f Breese et al (1989) shows firing density decreasing in a Gaussian fashion as normalised distance to the border o f fields increases. The shape and orientation o f place fields appears to be a function o f the environment the fields are found in. For instance, crescent shaped fields may often be found in cylinders but are rare in square and rectangular boxes, where the long axis o f elongated fields is parallel to box walls (Muller et al, 87 esp. fig. 6 and p. 1943; O ’Keefe and Burgess, 1996; Hartley et al, 2000; this thesis’ dataset). The range o f peak firing (in spikes o f all kinds per second) assigned to CS cells clearly depends on parameters adopted to describe the place field and its peak (eg., most obviously, spatial bin size, smoothing levels) and on the degree o f recording time allotted to given cells, but various methods o f describing peak firing can show place fields with up to 30 Hz peaks. The firing rate outside the field, as is often seen with well isolated cells, can approach 0 Hz.
Multiple place fields o f single cells
A single CS cell may fire, with high spatial specificity, in more than one region o f a bounded environment. While it was previously thought that this could be an artefact o f the available recording technology’s inability to distinguish two or more CS signal sources, and while the quantitative extent o f multiple peaks may not be agreed upon
the existence o f the phenomenon itself is not in question (eg. O ’Keefe and Burgess, 1996).
Place fields and directionality
Early reports often emphasised direction-specificity as well as location-specificity for place cells (O ’Keefe and Dostrovsky, 1971; O ’Keefe, 1976; McNaughton et al, 1983). A place field may be described as directional if the cell fires only, or at a higher rate, during traversals o f the field in a particular range o f directions. For instance, during radial maze tasks, a cell might fire when the rat runs through a field on outward journeys only. Subsequent studies have addressed this issue and produced the
following consensus: in open field environments, place fields are usually not directional, but in environments or tasks where rat traversals are spatially restricted, by environmental constraints (eg. maze arms) or by reward-shaping, fields are often directional (eg. Markus et al, 1995; Muller et al, 1987; M uller et al, 1994; O ’Keefe and Recce, 1993; O ’Keefe and Burgess, 1996). Many tasks combine both types o f traversal restriction. In the 8 arm radial maze for instance, the arms are usually narrow, and an efficient approach to reward-consumption in the task is arguably inconsistent with loitering in parts o f arms in all directions. The study by Markus et al (1995) suggests that the directionality seen in place fields is prim arily atttributable to the r a t’s restricted behaviour, and only secondarily, where it exists, to environmental constraints. Markus et al reshaped rats who had originally been trained to traverse an open cylinder in a pseudo-random way, to traverse the cylinder in a stereotyped way to obtain reward at four fixed reward sites. The proportion o f directional fields grew from less than 20% to nearly 40%.
Place Fields and angular location
As discussed in the section on head direction cells, it is likely that the presubicular directional system exerts control over the angular location o f place fields. In some studies investigators take pains to exert experimental cue control over static background and other uncontrolled cues. Rats are shaped to use the experimenter- chosen cues/systems (external or ideothetic) to control their sense o f orientation, and conflict between different cues or sensory systems contributing to orientation sense is minimised. In such circumstances, when the experimentally-chosen cues are rotated by a given amount, place fields can be observed to rotate both in unison and by a similar amount to the controlled cues (Jeffery et al, 1997; Jeffery and O ’Keefe, 1999; M uller et al, 87; O ’Keefe and Conway, 1978; O ’Keefe and Speakman, 1987; Olton et al, 1978). This is a well established finding, but it does not imply that the relationship between the angular locations o f place fields is always preserved following rotations. Certainly, in situations in which cues and sensory systems are in conflict (eg. visual vs vestibular systems, single cue vs another single cue), a group o f individual place fields can rotate by dissimilar amounts around a centre. This has been seen in the dataset o f the thesis (not shown), but is complicated by the fact that the experiment was done in a square. A study relating to this issue was recently done by Fenton and M uller (2000). Having first established that 2 cue cards on the walls o f a standard cylindrical environment controlled the orientation o f the place fields in the cylinder, they altered the distance between these cards. Although this procedure causes translation as well as rotation o f place field centres, the result emerges that place fields proximal to the cards are rotated more than place fields distal to the cards.
A key problem for all o f these studies is that it is not clear to what extent the hippocampal unit firing is driven by, or independent of, the directional system. Few studies have attempted to find dissociations between, for instance, simultaneously recorded head direction cells in presubiculum and C A l place cells.
The multimodal sensorv nature o f the place cell phenomenon
An element o f the hippocampal cognitive map theory (O ’Keefe and Nadel, 1978) was that place field activity could be derived on the basis o f various sensory modalities. In principle, input pertinent to spatial mapping such as distance and orientation
information could be obtained and integrated on the basis o f several external (eg. visual, somatosensory, olfactory, auditory) and internal cue responsive systems (eg. self-motion, proprioceptive, vestibular). Can place field firing be established and maintained on the basis o f distinct and/or multiple additive sensory input? What mechanisms are involved? Are some sensory systems privileged in hippocampal place field firing?
Early work by O ’Keefe and others (O ’Keefe and Dostrovsky, 1971; O ’Keefe, 1976; O ’Keefe and Conway, 1978) suggested that place units were opportunistic with regard to sensory input. One o f the problems with these studies, for our purposes here o f showing distinct and additive sensory control, is that relevant information might be extracted from any one or more sensory system’s responses to a “single” cue, like a buzzer or a running fan (visual, somatosensory, auditory, and olfactory). Evidence from the probe trials reported in the O ’Keefe and Conway (1978) study, where cues shown to have stimulus control were systematically removed, suggested that visual cues might have had more valency for 3 o f 8 cells studied in this way. One clear
observation nevertheless was that a high proportion o f place fields could be
maintained in the same position in the environment in the dark, at least on the second occasion that lights were switched o ff (O ’Keefe and Dostrovsky, 1971; O ’Keefe,
1976). Recent work has further confirmed this (Quirk et al 1990; Markus et al, 1994), though the basic observation is clouded by issues relating to the rat’s previous
experience prior to dark trials (Quirk et al, 1990) - as is anticipated by O ’K eefe’s observation that some place fields returned only in the second dark condition.
These studies are generally not able to elucidate mechanistic processes o f place field firing. One problem with the probe trial approach is that an input shown to support maintenance o f a place field is not necessarily an input which singly or in
combination with other inputs established the place field initially. Conversely, it is hard to show definitively that place field maintenance following cue removal implies a switch from one modality set to another, since it is often unclear which sensory modality or combination was responsible for the initial establishment. A place cell which continues to fire in the dark on a second trial might be doing so on the basis o f self-motion information, a modality which may or may not have helped establish the field. There is not sufficient space to tease out these matters fully here. It is more appropriate to review another line o f evidence about multimodality which comes from studies where certain sensory modalities are purposely impaired by experimenters.
Hill and others (1979; reviewed in O ’Keefe 1979) attempted to eliminate entire sensory modalities from rats undergoing place cell testing. Animals were blinded or deafened or had their vibrissae removed or had their olfactory receptors destroyed. Place cells were found in all these animals, with no obvious differences from those
found in normal rats. Hill and Best (1981) deprived animals o f both vision and hearing: again, place cells were found in these animals. Recently, using quantitative approaches developed by Muller et al (1987) and Skaggs et al (1993), Save et al
(1998) compared place cell firing in rats which were surgically blinded one week after birth with that o f normal rats. A cylinder, with four different objects placed at the periphery, served as the testing environment. No statistically significant differences were found except that the fields o f blind rats were slightly more directional, and that blind rats’ place cells had lower firing rates. The size o f the blind rats’ place fields, and their locational informational content were similar (ie. respectively smaller and higher, but not significantly so.) Intriguingly, unlike place cells in normal rats “which fired from the first moment o f entry into the environment” (ie. at the start o f a given trial, but after previous training in that environment), place cells in blind rats did not fire until the rats had “made physical contact with at least one object” . This might suggest that, for these rats, which might be expected to rely more on ideothetic information, ideothetic information alone was insufficient to reinstantiate place field firing. It might have been very interesting to test firing in a condition where the objects were removed.
The question o f which sensory modalities dominate place cells may be misconceived if the goal is to establish some kind o f natural prepotency o f a given modality in the rat. At any rate, definitive data on this are lacking. The purpose o f this section has been to review the evidence leading us to the general conclusion that there is an inherent flexibility and multimodality in the establishment and maintenance o f place fields. This evidence is consistent with O ’Keefe and N adel’s view that the
hippocampus has been highly determined, by evolutionary processes, to construct
allocentric place fields. Detailed knowledge about mechanisms o f sensory integration must await further research.
The relative importance o f distal and proximal cues
Hebb (1949) reviewed evidence to suggest that the position habits o f mammals with lower processing capacity, such as the rodent, were set up more by distal than proximal cues. From some o f the studies examining the angular location o f place fields mentioned above, in which it is generally observed that distal cues exert control over local cues such as box odours and maze arm textures (eg. Jeffery et al, 1997; O ’Keefe and Conway, 1978; O ’Keefe and Speakman, 1987; Olton et al, 1978), this suggestion appears very reasonable in the context o f place field study.
Two recent studies have explicitly addressed these issues, though it should be noted that once again the primary variable under investigation was the angular location o f place fields (Cressant et al, 1997; 1999). Cressant et al (1997; 1999) placed sets o f objects (a wooden cone, plastic cylinder, and red wine bottle) in the Muller cylinder and found that, when centrally placed, the objects failed to control the angular
location o f place fields, while the same objects placed at the periphery o f the cylinder exerted strong stimulus control over angular location. (The first o f these studies gave very preliminary indications that the stability o f place cells was reduced in the centrally-placed objects condition.) The authors interpret these results by endorsing reasonable Hebb-like arguments about the reduced computational load o f using distal objects to create a stable reference frame. It can be argued that these studies are really, in effect, investigating cue control o f directional system cells, but only indirectly so, through hippocampal place firing. Other issues o f stimulus control relating perhaps
more to specifically hippocampal function, such as whether distal cues are more important in establishing place fields, and controlling place field reliability, specificity and so on, have not been tackled.
W hile undisputed, these observations may be limited in scope by the specific types o f tasks used, by the nature o f the proximal and distal cues used, and because proximal cues are not made sufficiently distinctive or salient by the experimenters. For
instance, if rats were to be overtrained in conditions where visual distal cues did not consistently predict reward sites, but local textural cues did, different results might be seen. It is perhaps mistaken to overgeneralise data about place field control by distal cues, and from behavioural experiments, into a rule along the lines o f “distal cues always dominate over proximal cues (or are prepotent)”.
Place fields and environmental scaling - size and number
Ignoring for the moment the “absolute size” o f place fields seen in cells in different parts o f the hippocampus (discussed below: next section), it can be asked if place fields increase in size in bigger environments. Restricted to this simple question, findings are in accord, the answer being that they do increase in area (Hartley et al, 2000; Muller and Kubie, 1987). Probing further into the issue, however, shows up discrepancies. Muller and Kubie (1987) found, once sampling biases were controlled for, that when a similarly-located place field was observed in two environments o f similar shape, but different size, the areas o f the two resulting place fields differed less in scale than did the areas o f the actual environments. Thus, “firing field area does not scale up equally to the scaling o f apparatus size” (Muller, 1996). Hartley et al (2000), analysing the data o f O ’Keefe and Burgess (1996), suggest a different view.
however. Hartley et al measured the area o f the portion o f each place field firing at 50% or more o f the peak rate. Although the dataset is comparatively small (28 place cells), and the issue was not statistically examined, field areas were well predicted by environmental area. The mean field in the small square was 109 em^, and the mean field in the large square, four times the area o f the small square, was 385 cm^. The average o f the two means in the rectangles, which were h alf the area o f the large square, and double the area o f the small square, was 206 cm^.
Place field extent within an environment - are there differences in the dorsal and ventral hippocampus?
How large are place fields? This is a question that is in need o f an agreed calibrative framework. Certainly, current answers depends on the quantitative measures chosen by individual laboratories. In what has become the standard environment for the field, the 51cm high, 76 cm diameter cylinder, Muller et al (1987) found that the average field area for place cells was on average 22% o f the 2-dimensional circular floor surface, taking 1 Hz as the cutoff point. They found a correlation between field size in pixels and log(cutoff) was -0.995. To give a flavour o f the relationship, using 8 Hz as the cutoff point gave an average field size o f 4% o f the circular floor area. In his recent review o f place cells, Muller (1996) reported an average field size o f 13% o f the apparatus, with a 1 Hz cutoff. The reduction (22 to 13%) may well reflect the use o f better recording techniques. The use o f an absolute value could allow noise to contribute overmuch to the measure. Moreover, as described below, place cell firing rates show excess variance and are not well understood. It is possible that place field maps autoscaled to peak firing rate are more appropriate for considering place field extent. With this measure, the cutoff point becomes a proportion o f maximum firing
rate. Hartley et al (2000) reanalysed the data in O ’Keefe and Burgess (1996) in this way, to consider the peak area where firing rate exceeded 50% o f the maximum, for a given cell. Expressing Hartlley et al’s figures as a proportion o f the four rectangular environments, peak areas occupied about 3 % o f the available surface. It should be noted however that each peak was considered as a separate field area, where a cell had more than one field. It is not clear how much larger the 3% would be if all the peaks form a single cell count towards one total field area for that cell, since each field is auto scaled.
It should be noted that to date experimenters have not used large testing environments. Virtually all place cell experiments have been conducted in
environments where the available plane for rat to walk on is no more than 1 .5 x 1 .5m, at most.
Perhaps more importantly, all these figures are based on cells recorded in the dorsal CA fields o f the hippocampus. The septal (anterodorsal) and temporal (lateroventral) poles o f the hippocampus differ, however, in connectivity and neurochemistry. Briefly, the septal pole receives more sensory input from the external environment than the temporal pole, while the temporal pole receives more input from the rat’s internal environment than the septal pole. With such differences in mind, Jung et al