Familiarity and Recollection

I enter a friend’s room and see on the wall a painting. At first I have the strange, wondering consciousness, “surely I have seen that before,” but when or how does not become clear. There only clings to the picture a sort of penumbra of familiarity,—when suddenly I exclaim: “I have it, it is a copy of part of one of the Fra Angelicos in the Florentine Academy—I recollect it there!” — William James, The Principles of Psychology (1890, p. 658)

Separating Familiarity from Recollection

As illustrated above, there are clear functional differences between familiarity and recollection.

Familiarity reflects quantitative memory strength information about an item, rather than qualitative or associative information about an event—as is the case with recollection.

William James differentiated between familiarity-based recognition and recollection, which for James reflected the “recall” or “bringing back into consciousness” of past experiences.

Aristotle thought that sensory information passed through the blood stream to the heart where the memory traces of these events were stored.

James in contrast, thought that “currents” passed from the sensory organs through the nerves to the brain and that memory reflected the creation and modification of neural associations in the brain.

Baddeley’s anecdote beautifully illustrates the different subjective experiences of familiarity and recollection. When an object or a person looks familiar, the source of that familiarity is ambiguous:

• Is it a long-lost friend, someone we’ve seen in passing, somebody famous?
• Or just a person who resembles someone we know well?

Perceiving someone or an object as familiar is not sufficient for recollection, rather one must retrieve information that links that person or object to a unique past experience so that the memory is “about” or is “attributed” to some specific past events.

What types of associative information are sufficient to support recollection?

Some have argued that recollection must support controlled or discriminative responding (Jacoby, 1991). Thus, if one remembers where or when the information was learned, then one can choose to use that information or withhold it depending on one’s intentions.

Others have argued that the most important aspect of recollection is the retrieval of information linking the memory to the person or to the self (Conway, 2005; Tulving, 1985).

So for example, William James argued that recollection has a certain warmth and intimacy, and it “must be dated to my past. In other words, I must think that I directly experienced its occurrence.”

Tulving used the term autonoetic consciousness to refer to this special type of self-awareness associated with remembering. This differentiates remembering Opens in new window from other types of memory that are associated with different forms of conscious experience, such as the sense of familiarity one might experience when encountering a familiar painting (i.e., noetic consciousness), or a variety of implicit forms of memory that may occur with little or no conscious awareness.

Recollection is a relatively slow retrieval process

One of the core assumptions of a number of behavioral dual-process models is that familiarity is faster than recollection (Atkinson & Juola, 1974; Jacoby, 1991; Mandler, 1980). This assumption is also consistent with neuroanatomical models that assert that familiarity is based on brain regions earlier in the processing stream than those regions supporting recollection (Aggleton & Brown, 1999).

A variety of empirical results support this assumption. For example, results from response deadline and response signal procedures indicate that familiarity is available earlier than recollection (Toth, 1996; Yonelinas & Jacoby, 1994).

In addition, studies of event-related potentials have indicated that the electro-physiological correlates of familiarity are observed earlier than those associated with recollection. It should be noted, however, that these methods indicate when the products of these processes are available, not necessarily when the subject will make use of this information.

So, for example, it is generally the case that remember responses are made more quickly than knowing responses are (e.g., Dewhurst & Conway, 1994). Thus, although familiarity is generally available earlier than recollection is, familiarity may be used as a basis for responding only once the subject determines that recollection has not been successful.

Recollection generally operates independently of familiarity

Most dual-process models assume that recollection and familiarity function independently (Mandler, 1980; Jacoby, 1991). In line with these models, the empirical evidence indicates that the two processes are independent under most standard test conditions, but there are test conditions under which this assumption does not appear to hold.

To determine whether recollection and familiarity are operating independently, it is necessary to assess whether they can be functionally dissociated. That is, if the two processes are fully independent, it should be possible to produce double dissociations (e.g., find variables that have selective effects on one process, and other variables that have selective effects on the other process).

The behavioral literature indicates that recollection and familiarity can be doubly dissociated by a variety of manipulations. For example, response-speeding and dividing attention during time of test reduce recollection but leave familiarity unaffected (e.g., Gruppuso, Lindsay, & Kelley, 1997).

Conversely, relaxing response criterion and increasing processing fluency lead to changes in familiarity, but they have little effect on recollection.

Similarly, forgetting Opens in new window over intermediate retention intervals selectively influences familiarity, and changing the presentation modality of words between study and test also appears to selectively affect familiarity (Gregg & Gardiner, 1994).

Another line of evidence indicating that the two processes operate independently comes from the observation that the results of the process-estimation methods are verified by the results from the task-dissociation methods.

The estimation methods all assume that recollection and familiarity are independent, whereas the task-dissociation methods do not make any explicit assumption about how the two processes are related.

The fact that the task-dissociation methods verify the results of the estimation methods suggests that the assumptions underlying the estimation methods were not violated.

Finally, event-related potentials correlates of recollection and familiarity are found to be spatially and topographically distinct, such that recollection is associated with a late positive component, whereas familiarity has been associated with an earlier bifrontal component (Rugg & Yonelinas, 2003).

Although recollection and familiarity generally do appear to operate independently, there may be special cases in which they do not operate independently. For example, in studies in which performance levels are extremely high, the estimation methods produce patterns of results that are inconsistent with the results from studies in which such ceiling effects were avoided.

For example, most studies of aging indicate that healthy aging disrupts recollection but not familiarity. However, in studies in which estimates of recollection are extremely high (e.g., studies in which more than 60% of the studied items are recollected, and thus, overall recognition performance is close to ceiling), aging appears to disrupt both recollection and familiarity (Yonelinas, 2002).

Although these ceiling effects may simply reflect a methodological limitation associated with the estimation methods, it is possible that when performance levels are extremely high, the two processes no longer operate independently.

For example, if subjects remember a large proportion of the test items, and thus find only a few items to be familiar in the absence of recollection, this may affect how subjects map their subjective experiences onto remember and know responses, or how likely they are to endorse nonrecollected items in source recognition tests, and thus the two processes may no longer independently contribute to performance.

In either case, the important methodological implication is that one needs to be careful to avoid ceiling effects when examining recollection and familiarity.