Yonelinas, A. P. (2002). The Nature of Recollection and Familiarity: A review of 30 Years of Research. Journal of Memory and Language, 46, 441-517.

The purpose of this article is to explore recollection and familiarity by reviewing dual process models and looking at the empirical evidence provided for the models.

Dual process models:

Atkinson: Familiarity is a process well described by signal detection theory and supports memory for perceptual information. It reflects the activation of nodes where a node represents a word or object. Recollection occurs when an item falls between the set criteria for new items and old items and supports memory or semantic info.

Neuroanatomical Model: The most recent models suggest that hippocampal region is responsible for recollection, and surrounding structures in the medial and inferior temporal lobe are important for familiarity.

Mandler Model: Familiarity is assumed to support recognition memory judgments and performance on implicit memory where intraitem information is retrieved. Recollection is a process which that supports recognition and recall performance where interitem information is retrieved.

Jacoby: Recollection is an analytic consciously controlled process, where familiarity is automatic. The two processes are thought to operate in parallel.

Tulving: Similar to Atkinson model, but involves episodic memory ( recollection) and semantic memory (familiarity in absence of remembering.) Episodic memory stores personally experienced items and the semantic system store general knowledge about the world.

Yonelinas model: Familiarity reflects quantitative memory strength in a manner similar to signal detection theory. Recollection reflects the retrieval of qualitative information about a previous event.

Measurement methods:

Task dissociation methods: These aim to identify a task or test condition that isolates one of the two processes.

-Response speed methods. (response time method, response deadline method, and speed accuracy trade off). Because we agree that familiarity is expected to be faster than recollection, these should separate recollection and familiarity.

-Recall/Recognition methods: Assumes that recall relies more heavily on recollection than does recognition. So, when certain variables have a larger effect on recall than recognition, we can assume that they have a larger effect on recollection than familiarity. The method is limited because the two test conditions have different retrieval cues and responses, and because of scaling differences.

-Item/associative recognition methods: Tests of associative recognition (item-item associations, item-context associations, item-feature associations) measure recollection because it is assumed to reflect the retrieval of qualitative information. Familiarity only reflects the memory for studied or unstudied words, but is less useful in associative recognition tasks.

A limitation of the task dissociation methods is that they provide imprecise estimates of recollection or familiarity and some patterns of results will be ambiguous.

Process-Estimation methods: used to derive parameter estimates representing the contribution of recollection and familiarity to overall performance.

-Process dissociation procedure: Recollection and Familiarity parameters are estimated by using Inclusion and Exclusion performance. Inclusion instructions guide subjects to respond "yes" if the item was studied in either study list. (Recollection and familiarity can be used in inclusion) Exclusion instructions guide subjects to respond "yes" only if the item was studied in one of the test conditions. This procedure assumes that recollection and familiarity are two independent processes, and it should be possible to find variable that affect recollection without affecting familiarity.

-Remember/Know: Remembering (recollecting) and knowing (familiarity in the absence of recognition) are judgments made by the subjects about the basis of their memory for the items. This measurement technique assumes that the two processes are mutually exclusive which is inconsistent with all the other dual-process models. However, the independence remember/know procedure (IRK) provides estimates from remember/know judgments based on an assumption of independence.

-ROC: Estimates of the contribution of recollection and familiarity to recognition performance are made by examining the effect of different response criterion on hits and false alarms. Subjects rate their confidence of their yes/no recognition responses and then plot the hits against the false alarms as a function of response confidence. An advantage of this method is that it only requires one recognition test conditions. A limitation is a large number of observations.

Empirical findings

Encoding Manipulations:

Levels of processing: deep processing (processing the meaning) leads to an increase in recollection and a smaller increase in familiarity that shallow processing (processing the perceptual aspects) pg. 492. In all the 17 process estimation experiments conducted, semantic processing led to an increase in recollection.

Generation generating a word at a time of test is compared to reading a word. Generation compared to reading leads to a large increase in recollection and a smaller increase in familiarity. All 11 process-estimation experiments that examined generation showed the same pattern of results except for one. Pg 494

Divided attention during test: Requiring subjects to conduct a concurrent task while encoding the study items reduces recollection. Dividing attention has large disruptive effects of recall and smaller disruptive effects on recognition. So, recollection is more attention demanding than familiarity (Fig. 2, pg 494, 495)

Study duration: An increase in study duration leads to comparable increases in recollection and familiarity (Fig. 2, pg 495, 496)

Benzodiazepines: administered at time of encoding lead to a form of amnesia that that has a greater reduction in recollection than familiarity (Fig 2, pg 497)

Retrieval Manipulations:

Retrieval time: Studies of speeded recognition indicate that familiarity is available earlier than recollection, and nonspeeded tests indicate that familiarity typically contributes to performance earlier than recollection. In other words, early on familiarity is better than recollection, but as time passes, recollection improves and familiarity doesn’t change (Fig 3, pg 497)

Divided attention at test: Disrupts recollection but doesn’t influence familiarity. Divided attention at time of retrieval affects recall (recollection) but not recognition (familiarity). (Fig 3, pg 498)

Perceptual matching: Changing modalities of words leads to decrease in familiarity but no effect on recollection. For nonverbal materials, changing perceptual characteristics between study and test can decrease recollection and familiarity (Fig 3, pg 498)

Forgetting rates: For immediate delays, familiarity decreases and recollection doesn’t, and for long-term delays they both decrease (Fig 3, pg 499)

Fluency manipulations: Manipulations designed to increase the processing fluency of test items lead to an increase in familiarity based recognition responses, and recollection based responses are unaffected (Fig 3, pg 500)

False recognition: Items are falsely recognized on the basis of familiarity more often than they are falsely recollected, except when new items are highly associated to studied items (pg 501)

Shifting response criterion: Increasing subjects tendency to accept items as having been studied, has a large increase in the probability that items will be accepted on basis of familiarity, but not on recollection. (Fig 3, pg 501)

AND figure 4 says recollection is not described as a signal detection process because recollection estimates remain constant as false alarms increase instead of forming the guassian like curves consistent with the signal detection theory.

Stimulus variables

Word frequency: Recollection and familiarity are greater for low than high frequency words and the frequency effects are much larger for recollection than for familiarity, BUT sometimes there is a high frequency advantage for recollection. This probably happens because relationships will be formed between the test items making them easier to recall.

Memory for new items and new associations: Recollection can support memory for novel items and novel associations. Familiarity can support memory for novel items, but it can only support memory for new associations under some conditions. This goes back to the fact that associative recognition is based on recollection and familiarity does not benefit performance on these tasks.

Special populations

Amnesia: Damage that includes the hippocampus and surrounding temporal lobe disrupts both recollection and familiarity, but generally has a larger effect on recollection. (pg 504)

Aging Normal aging disrupts recollection but leaves familiarity largely unaffected.

Frontal lobe lesions Studies of recognition, recall, and associative recognition suggest that damage to regions in the frontal lobe disrupts recollection but leaves familiarity relatively unaffected.


ERPs During encoding, items that are associated with a widespread positive going ERP deflection are later recollected more than recognized on basis of familiarity. During retrieval, familiarity based responses are related to an early frontal scalp positivity, and recollection is related to a later left-parietal positivity.

Neuroimaging During encoding, activation of in the prefrontal cortex, hippocampus, parahippocampal gyrus, and bordering fusiform gyrus is associated with recollection. During retrieval, activation in the prefrontal cortex, hippocampus, parahippocampal gyrus, and parietal lobe is associated with recollection, whereas somewhat distinct prefrontal activations are related to familiarity.

Characterizing Recollection and Familiarity

Processing speed: The fact that familiarity is faster than recollection can be supported by Atkinson (familiarity is completed before recollection) and Mandler, Jacoby, and Yonelinas (they operate in parallel) and as long as the variability of response times related to familiarity is great enough, recollection and familiarity may be accounted for by a unimodal RT distribution.

The Relationship between Recollection and Familiarity Evidence supports the fact that recollection and familiarity are independent by the fact that they can be doubly dissociated by retrieval manipulations, and the fact that task- dissociation methods verified the conclusions derived from the process-estimation methods

The signal detection/Threshold distinction Familiarity is sometimes formalized using signal detection theory, where recollection has been formalized as a threshold process.

The conceptual/perceptual distinction The results show some support for the claim that recollection reflects a conceptual or elaborative process, and familiarity reflects a more sensory or perceptual process

Controlled Automatic Distinction Jacoby believes (and Atkinson, Mandler, and Yonelinas aggress) that recollection is relatively controlled and familiarity is an automatic process.

Intermediate term vs. long term memory Familiarity should decrease more rapidly than recollection.

Can familiarity support novel learning? The results suggest that familiarity supports memory for new items such random shapes and nonwords, but does not generally support learning of new associations as seen in tests of recognition.

The relationship between familiarity and implicit memory There is a lot of research showing that familiarity is functionally dissociable from performance on perceptual implicit memory tasks.

Neuroanatomical substrates: The lesion and neuroimaging results provide strong evidence that recollection relies on the prefrontal cortex and the hippocampus, and the neural substrates of familiarity are less understood.

Assessing the measurement methods: There are a wide variety of measurements to assess recollection and familiarity. Most of the time, the assumptions are supported by empirical evidence, but interpreting methods on recall must be done carefully because these measure differ from those of recollection. Also, it is important to design studies to avoid ceiling effects.

University of Arkansas
Department of Psychology
Graduate Program in Experimental Psychology
Lampinen Lab
False Memory Reading Group
False Memory Reading Group Fall 2002