Rate of Extinction
We often use the extinction curves as a proxy for either the magnitude of the CTA (i.e. the stronger the CTA, the longer it takes the CTA to , see many of our dose respnse curves) or the persistence of the CTA (i.e. two groups might start with the same initial magnitude of CTA , but one CTA might extinguish much slower than another (see Oberbeck data on okadaic acid treatment).
We are often loose in our definition of extinction rate, or when extinction has occured, however. Operationally, we define extinction to have occured when the average group preference is not significantly different the average preference of a sham control group (which often has a saccharin preference of >0.95 ± 0.05). With this measure, the "rate of extinction" is usually reported in our papers as the day when extinction first occured for the group.
There are a several problems with this approach.
- to define the lack of aversion based on a "lack of significant difference" with the control is to prove the null hypothesis, (as Professors B and M Licht delight in pointing out).
- it fails to provide a measure of variance within the CTA group; instead of getting a date of extinction for each member of the group, we instead only know when the group as a whole "extinguished".
- we get a single terminal number, but we do not capture any of the dynamics of what is presumably a continuous rate of extinction. Indeed, an animal might have a variable rate of extinction: slow at first, but speeding up after several days.
- there is no way to analyze the extinction rate with any finer granularity than a single 24-h 2-bottle test. How can we measure extinction within a single 24-h period, for example?
There are a several solutions to this problem. Hopefully I can document relevant examples of each approach.
- One is to bin the individual rats into different subgroups which have clearly distinct rates or patterns of extinction. The famous paper of Spector and Smith takes this approach, identifying groups which rapidly extinguished, slowly and gradually extinguished, extinguished rapidly but only after a delay of several days, or those which did not extinguish during the course of the experiment.
- Fit curves to the individual preference data. This was done very often in the days of skinnerian operant conditioning, e.g. with key pecks in pigeons undergoing extinction.
- Use absolute magnitude of saccharin intake. Problems: how to take into account accompanying water intake (or required necessity of fluid intake); advantage able to go above a preference of 1.0, to saccharin intakes greater than baseline water intake.
- Compare the individual's intake to an absolute value, rather than comparing the average of the group to the sham control group. Could look at copious historical data, and decide that e.g. "0.90" is the threshold for returning to an "unconditioned preference" for saccharin over water. Alternatively, one could use an absolute volume of saccharin, eg. saccharin intake which is greater than the rat's water intake during baseline would constitute recovery. The big advantage of this approach is that a specific date of extinction can be assigned to each animal, and thus an average date of extinction can be assigned to the group for statistical purposes.
(Note that there are some who would hold that as soon as saccharin > water, there is a preference for saccharin; however, even in this case the saccharin intake or preference can be significantly lower than that of the control group, and I consider a "significantly lower preference" to be conceptually an "aversion".)
Extinction during single vs. two bottle tests, restricted vs. unrestricted.
(I wonder if you could keep an animal covertly hydrated (saline drip into the stomach?) while giving access to e.g. saccharin. This would de-couple drinking for osmotic and volumetric purposes from drinking saccharin as a taste-stimulated behavior, subject to acquired aversions and preferences. In effect, drinking when hydration is "clamped" and accomplished in parallel might be a form of sham-drinking, decoupled from postingestive effects. Need to look at literature on this, e.g. Na appetite, IV feeding, etc.)