Let’s grant that behavioral loyalty is what pays the bills, but that attitudinal loyalty is also important, especially when it can be used as an indicator of higher behavioral loyalty. I think that’s the general, if not unanimous, conclusion of the discussion on this topic.
And when it is positioned as a straight yes-or-no proposition, the concept of customer loyalty, as a behavior, is relatively easy. A magazine subscriber who elects to renew her subscription at the end of the first year is engaging in loyal behavior.

However, the real world is rarely described adequately in strict yes-or-no terms. If the magazine subscriber renews her subscription again in the third year, and perhaps again in the fourth, fifth, and sixth years, doesn’t her behavior exhibit a greater and greater degree of “loyalty?” In other words, loyalty is not simply a yes-or-no proposition at all, but is a matter of degree.
And it can get more complicated than that. Consider a new car buyer. The owner of a Brand A car who buys another Brand A when he retires the first one would be said to be loyal, of course, but what about when he chooses to use his dealer’s service area, or to get his car financed from Brand A’s financial services division? Or what if he owns two cars, but only one of them is Brand A?
The “loyalty” concept is equally difficult to deal with in other categories – most categories, actually. Because most business categories are not simple subscription businesses. If a breakfast cereal consumer buys one box per month of Brand B for her family, and then begins buying two boxes a month, does this mean she is twice as loyal as before? What if she also went from buying one box a month of Brand C to buying three boxes a month? Would that mean she is LESS loyal to Brand B? (Note that we are still describing her behavior, here, even though we might be inferring her attitudes.)
The fact is, a much more useful concept than “loyalty,” when thinking about desirable customer behaviors, is probably “lifetime value.” The net present value of the expected stream of future profits attributable to a customer is a much more rigorous and useful variable, simply because it is a vector: it has both a direction AND a magnitude. In its ideal state (a state that can never actually be measured precisely, of course), lifetime value would capture all the various behaviors and activities of a customer that have any bearing at all on the enterprise’s profit from that customer.