Excuse me for starting a new post rather than appending to Howard's previous post. After thirteen comments it seems more convenient and better focused.
Howard's post calls dna a kind of memory. I agree that dna could be thought of as a sort of "species memory" for certain purposes, but I don't like this in a more general sense when one looks at the time factor. A current strand of dna is the result of multiple copying errors and mutations that have occurred in the past. Just because something is a RESULT of occurrences in the past doesn't mean that it is a RECORD of that past. I would prefer another approach based upon speculation about evolution.
The closest thing that I can think of to memory at a cellular level is "tropism," which may be the ancestor of true memory. For example single cells may be attracted to sunlight and away from shadow. They are wired to do this based upon natural selection and the nature of their food source. This sort of "pre-memory" or "proto-memory" records the fact that the more food is and has been associated with the light. It isn't true memory, because although it has both permanence and changeability, it isn't addressable. It's more like the set point on a thermostat. Although circumstances can cause a change in the setpoint, the current setting tells us nothing about the history of thermostat's settings.
To better understand the evolution of memory we need to ask why memory might be of value. In the case of photo tropism I'll create an imaginary situation. Suppose that the food versus light intensity relation was also a function of light frequency. A cell might evolve which had different receptors for different frequencies. Each of these receptors would have a different intensity setpoint. Hence, we have stored the information of the best intensity for the cell to propel itself toward, depending on the frequency of the light. The optimal intensity is therefore "addressable" in the sense that only that sensor is activated which corresponds to the particular frequency in the environment. A mechanical analog would be a combination thermostat and humidistat. Suppose we prefer various temperatures in a room depending on the humidity. Imagine that the humidistat selects a different thermostat with a different temperature setpoint depending upon the value of the humidity. We now have true memory. The array of thermostats contains all the temperatures existant and the humidistat makes them addressable according to the humidity. In effect saying, "I remember that when the humidity is x% you like the temperature to be y degrees." If there is any analog to this at the dna level, I believe it would be at the level of expression. A change in environmental conditions in the womb may cause the expression of certain genes to be encouraged or discouraged. Hence, the very same set of genes may produce different results, because of the inhibitors or promoters attached to the dna strand. Those inhibitor and promoter particles (I've forgotten their names) might be thought of as memory.
To summarize might we say that the existence of memory requires that an action shall be taken or not taken, depending upon a stored variable which is addressed by an environmental variable. At a macro human level this would be that a piece of information comes to consciousness from among many possibilities depending upon whether or not one finds the right access path or phrases the right question.
With respect -Joel