HHS Public Access Author manuscript Author Manuscript

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01. Published in final edited form as: J Exp Psychol Learn Mem Cogn. 2016 January ; 42(1): 114–126. doi:10.1037/xlm0000170.

Two Are Not Better Than One: Combining Unitization and Relational Encoding Strategies Hsiao-Wei Tu and Rachel A. Diana Virginia Polytechnic Institute and State University

Abstract Author Manuscript Author Manuscript

In recognition memory, recollection is defined as retrieval of the context associated with an event, whereas familiarity is defined as retrieval based on item strength alone. Recent studies have shown that conventional recollection-based tasks, in which context details are manipulated for source memory assessment at test, can also rely on familiarity when context information is “unitized” with the relevant item information at encoding. Unlike naturalistic episodic memories that include many context details encoded in different ways simultaneously, previous studies have focused on unitization and its effect on the recognition of a single context detail. To further understand how various encoding strategies operate on item and context representations, we independently assigned unitization and relational association to two context details (size and color) of each item and tested the contribution of recollection and familiarity to source recognition of each detail. The influence of familiarity on retrieval of each context detail was compared as a function of the encoding strategy used for each detail. Receiver operating characteristic curves suggested that the unitization effect was not additive and that similar levels of familiarity occurred for one or multiple details when unitization was the only strategy applied during encoding. On the other hand, a detrimental effect was found when relational encoding and unitization were simultaneously applied to one item such that a salient non-unitized context detail interfered with the effortful processing required to unitize an accompanying context detail. However, this detrimental effect was not reciprocal and possibly dependent on the nature of individual context details.

Keywords unitization; source recognition; familiarity; context information; ROC curves

Author Manuscript

Imagine that you are attending a wedding. The context around you includes details such as the smell of the flowers, the sound of the music, and the feel of the folding chair in which you sit. The attentional focus of the event, the item, might be the bride as she walks up the aisle. How are the context details of the event encoded? Are they integrated with one another into a cohesive unit that is then bound with the item? Is each detail linked

Correspondence concerning this article should be addressed to: Hsiao-Wei Tu, North End Center, Suite 4300, 300 Turner Street NW, Blacksburg, VA 24061. Phone: 1-540-231-8131. Fax: 1-540-231-3380. [email protected]. Hsiao-Wei Tu and Rachel A. Diana, Psychology Department, Virginia Polytechnic Institute and State University. Hsiao-Wei Tu is now in the Neuroscience Division of the Academy of Integrated Science, Virginia Polytechnic Institute and State University.

Tu and Diana

Page 2

Author Manuscript

individually to the item? Are some details encoded as features of the item? If so, how do these “item features” relate to the remaining context details? The current study investigated the encoding of multiple context details within a single event. We examined the effect of consistent encoding strategies applied to all context details as compared to different encoding strategies applied to individual context details and how these encoding strategies affect retrieval.

Author Manuscript

Recognition memory, the ability to identify that a stimulus was previously encountered, can be supported by two processes: familiarity and recollection. Familiarity is defined as retrieval based on stimulus strength alone, whereas recollection is defined as retrieval of both the stimulus and other information that was associated with it during encoding (e.g., Mandler, 1980; Rugg & Yonelinas, 2003; Wixted, 2007; Yonelinas, 2002). Either of these retrieval processes can be more or less diagnostic for a given memory task. Typically, memory for the relationship between two random items strongly relies on recollection, but memory for each single item that comprises the pairs can be retrieved by both familiarity and recollection (Hockley & Consoli, 1999; Turriziani, Fadda, Caltagirone, & Carlesimo, 2004). Similarly, memory for arbitrary associations between an item and its context details (e.g., the background color of an object) is highly supported by recollection, but memory for the item itself can be recognized based on both familiarity and recollection (Davachi, 2006; Macken, 2002; Yonelinas, 2001).

Author Manuscript

Recently, two lines of research have challenged the assumption that context or associative retrieval must always be based on recollection. Perceptual binding produced at encoding may change how familiarity and recollection contribute to source recognition (Boywitt & Meiser, 2012a). Specifically, “intrinsic” features, which are visually combined with the item, are automatically encoded with the item and thus can be recognized based on familiarity, whereas “extrinsic” features, which are visually separate from the item, require intentional processing as they are not part of the item and therefore are retrieved mainly by recollection (Boywitt & Meiser, 2012b; Mulligan, 2011; Zimmer & Ecker, 2010). Returning to the wedding event, the bride’s dress is visually combined with the item (the bride herself) and thus is likely to be automatically encoded as part of the item information.

Author Manuscript

In addition to perceptual processing, different semantic encoding processes affect the nature of an event’s representation and therefore change the contribution of familiarity and recollection at retrieval. In particular, instead of making relational connections, which are by nature arbitrary and flexible, two items can sometimes be encoded as a single, unified item or context can sometimes be encoded as a feature of an item. This process has been labeled as “unitization” (e.g., Graf & Schacter, 1989; Yonelinas, Kroll, Dobbins, & Soltani, 1999). Retrieval of unitized information can rely on familiarity, rather than exclusively recollection that necessarily supports retrieval of non-unitized, relationally encoded information (Diana, Yonelinas, & Ranganath, 2008; Giovanello, Keane, & Verfaellie, 2006). In the wedding event, the bride and her partner might be effortfully combined into a new, single item unit. For example, rather than encoding that you spent the day at Pam’s wedding, you might encode that you spent the day at Pam & Pat’s wedding. Pam & Pat in this case would be a new item that is encoded holistically as a single representation. The key difference between perceptual binding and unitization is that perceptually bound associations are based purely

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 3

Author Manuscript

on the co-occurrence of the item and the context, whereas unitized associations are based on the re-conceptualized identity of the item and the context.

Author Manuscript

The effects of unitization have been examined in many studies over the past decade (e.g., Diana et al., 2008; Ecker, Zimmer, & Groh-Bordin, 2007a; Haskins, Yonelinas, Quamme, & Ranganath, 2008; Watson, Wilding, & Graham, 2012). Evidence from behavioral, patient, and fMRI studies has converged to support the notion that unitization of item and context at encoding promotes the engagement of familiarity in retrieving source information. For example, in a word pair recognition test (Giovanello et al., 2006), participants reported that they used familiarity to recognize word pairs more often when those words formed a compound word (e.g., black-mail) than when the words in the pair were unrelated (e.g. surgeon-arrow). This result suggests that unitizing two objects into a single compound stimulus increases the contribution of familiarity, although associative recognition without unitization relies primarily on recollection. In another study, Quamme, Yonelinas, and Norman (2007) tested amnesic patients, who have impaired recollection but spared familiarity, using word pairs. In the unitized condition, a definition was given that made one word a modifier for the other within a single new concept, whereas in the non-unitized condition, a sentence was given that included the two words as separate lexical units. Recognition accuracy for recollection-impaired patients was better for unitized than nonunitized stimuli. Given that these patients primarily use familiarity to make memory judgments, these results indicate that unitization increases the contribution of familiarity to associative memory as compared to word pairs that are not unitized.

Author Manuscript

In addition to “combining” a pair of items as a single unit, unitization can also be used to create a new representation that includes both an item and its associated context details. Staresina and Davachi (2006) induced unitization of a study item with its context by asking their participants to imagine the study item as if it were the same color as the background on which it was superimposed and report whether this imagined item was plausible. Functional brain imaging showed that perirhinal cortex, typically involved in processing the strength or familiarity of individual items/objects (Davachi, 2006; Ranganath et al., 2004; Weis et al., 2004), was activated when source information of those unitized items was recognized, whereas parahippocampal cortex and the hippocampus, typically supporting recollection (Davachi, 2006; Ranganath et al., 2004; Rugg & Vilberg, 2013), were more activated when retrieving non-unitized contextual information. In a similar paradigm, Diana et al. (2008) found increased contributions from familiarity to receiver operating characteristics (ROCs) for unitized items and contexts as compared to non-unitized items and contexts.

Author Manuscript

In naturalistic events, many context details are encoded, such as the flowers and music at the wedding, and presumably these details can be relationally bound with the item (whether unitized or not) into a memory representation (Uncapher, Otten, & Rugg, 2006). However, previous studies of source recognition focused on either unitization or relational encoding within a single event and evaluated its effect on the contribution of familiarity and recollection at retrieval. Moreover, both memory models and empirical data showed that individual context details interact differently, depending on the memory retrieval process being used. Specifically, in conscious recollection, pieces of source information are stochastically related, whereas in familiarity-based recognition, multiple context details are

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 4

Author Manuscript

processed independently (Meiser & Bröder, 2002). In order to better understand how relational (non-unitized) encoding and unitization interact in the same scenario, we assigned these two encoding strategies to two context details (size and color) respectively in a paradigm that has previously been shown to increase familiarity for unitized background color (Diana et al., 2008).

Author Manuscript

In previous studies using this paradigm, participants in the unitized group were asked to form a mental image of the item in the same color as the background (e.g., a red violin), whereas those in the non-unitized group were asked to form a mental image of the item in an event with either a stop sign (red) or dollar bill (green). At test, all participants evaluated their level of confidence that a stimulus was studied with either a red background or a green background. Correct and incorrect source responses were plotted for each confidence level and the resulting data were fit to the dual-process signal detection (DPSD) model. The DPSD model hypothesizes that familiarity reflects a signal detection process and therefore generates more curvilinear ROC curves, whereas recollection operates as a threshold process (i.e., some items do not lead to retrieval of any relevant context information) and results in linear ROCs (Parks & Yonelinas, 2007; Yonelinas, 1999). Indeed, Diana et al. (2008) found more curvilinear ROCs for participants given unitized instructions than for those given relational encoding instructions. This result suggests that the unitized stimuli were more likely to be recognized using familiarity-based recognition.

Author Manuscript Author Manuscript

In the present study, a sentence was provided with each study word to simultaneously manipulate the encoding strategy used for color and size information and induce participants to create unitized and relational representations. The extent to which familiarity and recollection processes were engaged in recognition was estimated based on the shape of ROC curves (Yonelinas, 1997, 1999; Yonelinas & Parks, 2007). Although the unitization theory states that combining the two details into a single representation should allow retrieval of the association via familiarity (Diana et al., 2008; Graf & Schacter, 1989; Yonelinas et al., 1999), it does not propose any interactions between unitization of a single context detail and encoding of other context information in the same episode. Considering that retrieval of unitized and non-unitized source information independently activated different subregions in the medial temporal lobe (Davachi, 2006; Ranganath et al., 2004; Rugg & Vilberg, 2013; Weis et al., 2004), we predicted that the two encoding strategies can be applied independently. The involvement of familiarity in source memory for one detail should be dependent on the unitization condition of that detail alone and unaffected by the unitization condition of the second context detail. Specifically, ROC curves for font size judgments would be more curvilinear when font size was unitized than when it was not, regardless of whether color information was or was not unitized. In reverse, ROC curves for color judgments would be more curvilinear when background color was unitized, regardless of the encoding status of the size information. Alternatively, relational memory theory (Cohen & Eichenbaum, 1993) and the Binding of Item and Context theory (Diana, Yonelinas, & Ranganath, 2007) both argue that the hippocampus creates a holistic representation of a complete episode. Any context information unitized with the item would ultimately be bound with other separately encoded context information via the hippocampal binding process. Therefore, a unitized item and

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 5

Author Manuscript

context would interact during encoding processing with any additional context details that are relationally bound to the episode. If this is the case, unitization of either color or size might lead to increased familiarity for all components of the episode, and unitization of both color and size might further increase familiarity beyond the level found when only a single item is unitized.

Experiment 1

Author Manuscript

In Experiment 1, the background color (a visually extrinsic feature) and font size (a visually intrinsic feature) associated with each item were either unitized or not unitized at encoding, creating four conditions: Unitize Color (meaning size was not unitized), Unitize Size (meaning color was not unitized), Unitize Both, and Unitize Neither (Table 1). We expected that the semantic manipulation of unitization would be stronger than any effects of perceptual binding. Assuming that relational encoding and unitization, each operating on an individual context detail, have independent effects on source recognition, we predicted that the ROCs for each memory judgment (color or size) would be more curvilinear (indicating an increased contribution from familiarity) when that piece of information was unitized with the item than when that piece of information was not unitized with the item. Unitization of a second detail was not expected to influence the curvilinearity of the ROC curve for the first detail. Methods

Author Manuscript

Participants—Thirty-two students from Virginia Tech gave informed consent to participate in Experiment 1 for monetary reward or extra credit in psychology courses. Each participant completed all four conditions, but data from eight participants were not included in the analyses because their performance was lower than chance level (measured d′ ≤ 0 in any condition). Data including these participants produced the same pattern as did the analyses reported here.

Author Manuscript

Materials—Sixty words were used in each of the four conditions. Fifteen words were presented with each of the four combinations: either big (60 points) or small (20 points) font on either red or green background. Each word was accompanied by a sentence that described a scenario in which the given size and color were incorporated (see Table 1 for examples). When unitization was applied, the scenario described the size and/or color of the item itself. When relational encoding was applied, the scenario described the size and/or color of separate objects. That is, when background color was unitized (in the Unitize Color and Unitize Both conditions), the sentence provided a reason why the study item was red or green. When background color was not unitized (in the Unitize Size and Unitize Neither conditions), the sentence provided an association between the study item and a stop sign (red) or a dollar bill (green). Similarly, when font size was unitized (in the Unitize Size and Unitize Both conditions), the sentence provided a reason why the study item was big or small. When font size was not unitized (in the Unitize Color and Unitize Neither conditions), the sentence provided an association between the study item and an elephant (big) or a mouse (small).

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 6

Author Manuscript

Procedure—The study included four blocks of study and test phases that corresponded to the four conditions. The order in which these four conditions were presented was counterbalanced across participants. Each block started with instructions on the screen that informed the participant whether the sentence would provide reasons or make associations for color and size information, respectively. During the study phase, a trial began with a fixation cross for 1 sec, followed by a word and its corresponding sentence presented for a fixed 6 sec. Participants were told to study the given sentence, imagine the scene it described, and focus on the imagined scene to remember the item as well as its color and size, rather than incorporating additional strategies. Also within the 6 sec interval, participants were asked to rate the vividness of the mental image they created from 1 (very vivid) to 3 (not vivid). The 60 words in each study phase were presented in a random order.

Author Manuscript

In the four test phases, which immediately followed each of the four study lists, a trial started with a fixation cross on the screen for 1 sec, followed by simultaneous presentation of a word, a test question, and 6 confidence levels. All words appeared in the same font size (36 points) across trials on a black background in a random order (different than the order in the study phase). For each word, both color and size were tested individually but the order of the two questions was random. For each memory judgment, participants were asked to indicate their confidence on a scale from 1 to 6 with 1 representing the most confident “green” or “small” response, and 6 representing the most confident “red” or “big” response. Subjects had a maximum of 6 sec to respond.

Author Manuscript

Analysis—We evaluated overall performance by collapsing across confidence judgments to create measures of hits and false alarms. These values were used to calculate measured d′ (Z(hit rate) – Z(false alarm rate)) for each context detail in each participant to examine successful memory discrimination. Vividness ratings were compared across conditions by related-samples Friedman’s ANOVA by ranks. ROC curves were calculated for the background color and font size judgments separately in each condition. The degree of curvilinearity of each ROC was assessed by fitting the DPSD model to the observed ROCs. Familiarity estimates, which constitute the d′ parameter of the model, were the primary dependent variables in our study. Two repeated measures ANOVAs were conducted to compare color and size familiarity estimates separately across conditions. In both ANOVAs, the encoding strategy (unitized vs. non-unitized) of color was matched with the encoding strategy (unitized vs. non-unitized) of size in a complete 2-by-2 design. In addition, paired t tests were used to examine differences in familiarity estimates for size judgments and color judgments within each condition.

Author Manuscript

Recollection estimates were also derived from the DPSD model and reported as probabilities. In each condition, recollection estimates for green background and those for red background were averaged as color recollection estimates, whereas recollection estimates for small font size and those for big font size were averaged as size recollection estimates. The color and size recollection estimates were analyzed by two repeated measures ANOVAs respectively across conditions and by paired t tests within the same condition.

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 7

Results and Discussion

Author Manuscript

As summarized in Table 2, performance (measured d′) on color and size recognition was significantly above chance in every condition, ts > 8.42, ps < 0.001. There was no significant difference between color accuracy and size accuracy of the same word in all four conditions, ts < 1.52, ps > .13. Also, average vividness did not differ across conditions, χ2(3) = 5.84, p = .12, indicating that the words and encoding sentences used in this study were not biased toward any context detail. The DPSD model provided an excellent fit to the observed data in all four conditions with sum of squared errors ranging from 0.002–0.004.

Author Manuscript

Effect of encoding strategies on familiarity across conditions—Analysis of familiarity estimates for color showed a main effect of encoding strategy used for color information, F(1,23) = 4.34, p = .05, ηp = .16 (Figure 1A). That is, familiarity estimates for color recognition judgments were significantly higher when color was unitized (in the Unitize Both and Unitize Color conditions) than when color was not unitized (in the Unitize Size and Unitize Neither conditions). In contrast, there was no main effect of size encoding strategy such that how size information was encoded did not affect the amount of familiarity involved in color recognition judgments, F(1,23) = 1.70, p = .21, ηp = .07. However, a significant interaction was found between encoding strategy and context type, F(1,23) = 7.59, p = .01, ηp = .25. Further analysis of this interaction showed that when color was unitized, the contribution of familiarity to color recognition was greater when size was also unitized than when size was not unitized, Unitize Both vs. Unitize Color: t(23) = 2.80, p = . 01, d = .57. But when color was not unitized, the contribution of familiarity to color recognition was equally low regardless of whether size was or was not unitized, Unitize Size vs. Unitize Neither: t(23) = 1.27, p = .22, d = .26.

Author Manuscript

Figure 1B shows the ROCs of color recognition in probability space. The ROC curve in the Unitize Both condition was more curvilinear than those in the other three conditions, which is reflective of the higher familiarity estimates when a unitizing strategy was used to encode both color and size information than when it was only used on one detail or not at all. Contrary to our prediction, the ROC curve in the Unitize Color condition did not show an increase in curvilinearity but instead overlapped with the ROCs in conditions in which color was not unitized (i.e., the Unitize Size and Unitize Neither conditions).

Author Manuscript

Analysis of familiarity estimates for size showed a main effect of encoding strategy used for size information, F(1,23) = 10.17, p = .004, ηp = .31 (Figure 1C). That is, familiarity estimates for size recognition were significantly higher when size was unitized (in the Unitize Both and Unitize Size conditions) than when size was not unitized (in the Unitize Color and Unitize Neither conditions). We also found a main effect of color encoding strategy such that familiarity estimates for size recognition were significantly higher when color was unitized (in the Unitize Both and Unitize Color conditions) than when color was not unitized (in the Unitize Size and Unitize Neither conditions), F(1,23) = 5.06, p = .03, ηp = .18. There was no significant interaction between encoding strategy and context type with respect to familiarity estimates for size judgments, F(1,23) = 1.44, p = .24, ηp = .06. Paired t tests revealed that, when size was unitized, unitization of color significantly increased the involvement of familiarity in size recognition, Unitize Both vs. Unitize Size: t(23) = 2.09, p

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 8

Author Manuscript

= .05, d = .43. However, when size was not unitized, unitization of color did not affect the involvement of familiarity in size recognition, Unitize Color vs. Unitize Neither: t(23) = 1.01, p = .33, d = .21. Figure 1D shows the ROCs of size recognition in probability space. Consistent with the ROCs of color recognition, the ROC curve in the Unitize Both condition was more curvilinear than those in the other three conditions, indicating higher familiarity estimates when unitization was the only encoding strategy adopted in an event. It also showed that although size was unitized in both the Unitize Size and Unitize Both conditions, size recognition was more heavily supported by familiarity when color was also unitized. Moreover, when size was not unitized, the ROC curve was more curvilinear in Unitize Color condition than in the Unitize Neither condition, suggesting that unitizing color facilitated the use of familiarity in size recognition.

Author Manuscript

Taken together, the results suggest that relational encoding and unitization do not seem to operate independently in the same scenario as we hypothesized. Specifically, when unitization was the only encoding strategy applied to multiple details in an event, a greater contribution from familiarity to source recognition was observed, compared to the familiarity estimates in conditions where the two encoding strategies were used simultaneously. There are two potential explanations for this pattern of results. One interpretation is that unitization of a single detail may facilitate the application of unitization on a second detail. Another possibility is that the process of unitizing a single detail could be hindered by making relational association of a second salient detail.

Author Manuscript

Effect of encoding strategies on familiarity within conditions—Here we directly compared the effects of different encoding strategies on color judgments and size judgments within each of the four conditions. In the Unitize Both condition when unitization was the only strategy used for both details, familiarity estimates were high but did not differ significantly between color recognition (M = .83) and size recognition (M = 1.08), t(23) = 1.39, p = .18, d = .28. Similarly, in the Unitize Neither condition when relational encoding was the only strategy used for both details, familiarity estimates were low but did not differ significantly between color recognition (M = .55) and size recognition (M = .47), t(23) = .81, p = .43, d = .17.

Author Manuscript

In conditions when the two encoding strategies were applied simultaneously in the same scenario, we predicted that only the unitized context detail could be retrieved by familiarity. In the Unitize Size condition, familiarity estimates for size (M = .76) were significantly higher than those for color (M = .39), t(23) = 2.24, p = .04, d = .46. However, in the Unitize Color condition, familiarity estimates for color (M = .44) were not significantly higher than those for size (M = .57), t(23) = 1.54, p = .14, d = .31. The pattern of results for familiarity estimates suggests some asymmetry in the effects of unitization on size and color. In particular, processing non-unitized size information seems to affect the unitization of color information in episodic memory more than the processing of non-unitized color information affects the unitization of size information.

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 9

Author Manuscript

Effect of visual presentation on familiarity estimates—When data were collapsed across different conditions, familiarity estimates for size (M = .72), which was presented as a visually intrinsic detail, were significantly higher than those for color (M = .55), t(95) = 2.38, p = .02, d = .24, which was presented as a visually extrinsic detail. This result suggests that unitization of font size may have been perceptually facilitated and therefore familiarity provided a larger contribution to size retrieval than color retrieval.

Author Manuscript

Recollection estimates—Although we did not have predictions about the recollection estimates in this experiment, Table 2 summarizes recollection estimates in each condition. There were no significant differences between recollection estimates for color and size judgments in any of the four conditions, ts < 1.89, ps > .07. However, we found main effects of color encoding strategy, F(1,23) = 11.34, p = .003, ηp = .33, and size encoding strategy, F(1,23) = 14.55, p = .001, ηp = .39, in color recognition such that recollection estimates for color were significantly higher when either color or size was unitized, but there was no significant interaction, F(1,23) = 1.18, p = .29, ηp = .05. Similarly, there were main effects of color encoding strategy, F(1,23) = 9.56, p = .01, ηp = .29, and size encoding strategy, F(1,23) = 8.38, p = .01, ηp = .27, in size recognition such that recollection estimates for size were significantly higher when either of the details was unitized, but there was no significant interaction, F(1,23) = 3.27, p = .08, ηp = .13.

Experiment 2

Author Manuscript Author Manuscript

The findings in Experiment 1 have raised two questions. First, it is unclear whether the significantly higher contribution of familiarity to source recognition in the Unitize Both condition was due to an additive effect of unitizing two details at once, a detrimental effect of mixing different encoding strategies, or a combination of both. In contrast to an early study solely focusing on the unitization effect on source recognition (Diana et al., 2008), we did not find significantly higher contribution of familiarity to color recognition when background color was unitized with the study item during encoding in the Unitize Color condition. We attributed this unexpected result to the addition of a non-unitized size detail, which was the primary difference between the two experiments. We hypothesized that the process of associating salient size information with the item decreased the effect of unitization and thus the influence of familiarity on color judgments. Experiment 2 tested this hypothesis in a condition in which the size information was removed from the sentences, leaving background color as the only context information unitized with the study word. If relationally encoded size information affects the unitization of background color, we expected that familiarity estimates for color recognition would be higher in this “Unitize Color Only” condition in Experiment 2 than those in the Unitize Color condition in Experiment 1. The other puzzling result from Experiment 1 was that unitization may have asymmetric effects on color and size recognition. One explanation for the difference in results for color and size recognition may lie in the way that color and size were represented in the nonunitized size scenarios. When size was not unitized, the item was associated with a mouse for small font size or an elephant for big font size, whereas when color was not unitized, the item was associated with a dollar bill for green background or a stop sign for red

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 10

Author Manuscript

background. Dollar bills and stop signs are both inanimate objects that do not perform actions, whereas a mouse or elephant can often be the actor in a scene. When size was not unitized, our attention may have been drawn from the study item to the mouse or elephant due to its active role in the provided sentence. In Experiment 2, we divided the Unitize Color condition into two conditions, depending on whether mouse/elephant was active in the scenario being depicted by the encoding sentence. If the asymmetry between color and size recognition in Experiment 1 was due to the nature of the items in the non-unitized sentences, we would expect that memory for color should rely more heavily on familiarity when mouse/elephant is a passive object than when it is an active subject during encoding. Methods

Author Manuscript

Subjects—Twenty-nine students, who were not included in Experiment 1, participated in Experiment 2 in exchange for monetary reward or extra credit in psychology classes. Five participants were excluded in the analyses due to chance-level performance (measured d′ ≤ 0 in any condition). Data including these participants produced the same pattern as did the analyses report here. Materials—There were three conditions in the current study; each included 60 words and 60 corresponding encoding sentences. In the Color Only condition, all words appeared in the same font size (36 points). Half of them were presented on a red background and given a sentence that provided a reason for the study item to be red (e.g., “The DOUGH is red because it is mixed with strawberries.”), and the other half were presented on a green background and given a sentence that provided a reason for the study item to be green (e.g., “The BENCH is green because it is covered in pollen.”).

Author Manuscript

In both the Active Size and Passive Size conditions, color was always unitized. Each study word was accompanied by a sentence that described why the item was red or green (unitizing the color) and that associated the item with a mouse or an elephant. In the Active Size condition, a mouse or elephant performed an action in the scenario (e.g., “The TANK is green because a mouse destroyed the pump and algae started to grow.”), whereas in the Passive Size condition, a mouse or an elephant was only passively associated with the studied item in the scene (e.g., “The TIRE is red because the driver was drunk and ran over a mouse.”). Fifteen words were presented with each of the four context detail combinations: either big (60 points) or small (20 points) font on either red or green background.

Author Manuscript

Procedure—Each participant completed three study-test blocks that included the Active Size and Passive Size conditions in a random order, followed by the Color Only condition. The procedure used here was the same as in Experiment 1, except that participants were not given size information in the Color Only condition. Analysis—Performance (measured d′), ROC curves, familiarity and recollection estimates were calculated in the same way as in Experiment 1. Vividness ratings were compared between the Active Size and Passive Size conditions by related-samples Wilcoxon signedrank test. Paired t tests were used to examine differences in familiarity estimates and in recollection estimates for size judgments and color judgments both across and within the

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 11

Author Manuscript

Active Size and Passive Size conditions. In addition, we compared familiarity estimates for color recognition from: 1) the Color Only condition, in which there was no size information to remember and color was the only unitized context detail; 2) the Unitize Color condition in Experiment 1, in which size information was present but not unitized; and 3) the Unitize Both condition in Experiment 1, in which size information was present and also unitized with the item. Results and Discussion

Author Manuscript

Effect of relational encoding on unitization—The DPSD model provided an excellent fit to the observed data with an average sum of squared error of 0.001 in each condition. Figure 1A (triangle icon) shows that unitizing color in the absence of size information (Color Only, M = .88) led to a similar contribution from familiarity as did unitizing both color and size information at the same time in Experiment 1 (Unitize Both, M = .83), t(46) = .28, p = .78, d = .08. Moreover, familiarity estimates in the Color Only condition were significantly higher than that in the Unitize Color condition (M = .44) when color was unitized while additional association was made between the item and its size information, t(46) = 2.71, p = .01, d = .78. These results suggest that unitization of a second context detail does not promote unitization effect on the first detail but rather that relational encoding of a second context detail reduces the unitization effect on the first detail. The level of familiarity involved in recognition of a unitized context detail stays the same regardless of whether unitization was only applied to that detail or to a second detail at the same time. This conclusion is also reflected in the similarly curvelinear ROC curves of color recognition in the Color Only and Unitize Both conditions (Figure 1B, dotted line).

Author Manuscript

The results from these analyses do not support the hypothesis that using unitization again on an additional context detail facilitates familiarity-based recognition of the first unitized detail. Instead, our data suggest that combining relational encoding and unitization decreases the contribution of familiarity to retrieval of the unitized context detail. Effect of different non-unitized size presentations on source recognition— Overall, performance (measured d′) on color and size recognition was significantly above chance in both Active Size and Passive Size conditions, ts > 11.00, ps < 0.001. There was no significant difference between color accuracy and size accuracy of the same word in any condition, ts < .37, ps > .72. Also, average vividness did not differ across conditions, Z = . 34, p = .74, indicating that the words and encoding sentences were not biased between conditions (Table 2). Average sum of squared errors of model fit ranges from 0.001 to 0.003.

Author Manuscript

Contrary to our predictions, familiarity estimates of color did not differ, regardless of whether mouse/elephant was an active subject (Active Size, M = .52) or a passive object during encoding (Passive Size, M = .65), t(23) = 1.07, p = .30, d = .22) (Figure 2A). Familiarity contribution to size recognition also did not differ between conditions (Active Size, M = .72; Passive Size, M = .73), t(23) = .09, p = .93, d = .02 (Figure 2B). Additionally, within each condition, familiarity estimates for color recognition did not differ significantly than those for size recognition, ts < 1.27, ps > .22. These results suggest that the active or

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 12

Author Manuscript

passive nature of the non-unitized associations do not affect the familiarity estimates for the accompanying unitized detail. Recollection estimates—There were no significant differences between color and size recollection estimates in either the Active Size condition or the Passive Size condition, ts < 1.33, ps > .20 (Table 2). However, both color and size recollection estimates were significantly higher when mouse/elephant was the actor in the encoding sentence than when it was not, ts > 2.27, ps < .04.

Experiment 3

Author Manuscript

Experiment 2 identified the direction of the effect found in Experiment 1. That is, addition of a relationally-bound context detail reduced the contribution from familiarity to retrieval of a unitized detail. We found that memory for a single unitized detail produced a similar level of familiarity to the Unitized Both condition and significantly more familiarity than the Unitize Color condition, in which size was not unitized, in Experiment 1. We also hypothesized that the asymmetric unitization effect on color and size recognition in Experiment 1 may have been be caused by the way associations were built between the study item and its size information. However, results from Experiment 2 did not support this hypothesis. That is, whether size information was associated with the item through an active role or a passive object did not influence the contribution from familiarity to color recognition.

Author Manuscript

A second possible explanation for the difference between color and size in Experiment 1 was the method of perceptual presentation of the two details. Experiment 3 was designed to test whether visual presentations of context details affect the extent to which familiarity contributed to source recognition in addition to unitization, resulting in the asymmetric effect we observed. In this experiment, the same words and sentences used in Experiment 1 were used again in the four conditions, but size information was extrinsically presented as the size of a white box that surrounded each study word, whereas color information was intrinsically presented as the color of the study item itself. We expected the opposite asymmetry to what we observed in Experiment 1. In particular, relational processing of an intrinsic, non-unitized context detail (font color in this case) would disrupt the unitization of another extrinsic context detail (box size in this case). In other words, familiarity contribution to size recognition should be low in the Unitize Size condition because of the association made between the color information and the item. Predictions for other conditions stayed the same as those in Experiment 1. Familiarity should support color recognition more so than size recognition in the Unitize Color condition. Familiarity estimates should be high for both details in the Unitize Both condition and low for both details in the Unitize Neither condition.

Author Manuscript

Methods Participants—A distinct group of thirty-two students from Virginia Tech were recruited for Experiment 3 and received monetary reward or extra credit in psychology courses. Data from eight participants were not included in the analyses because their performance was

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 13

Author Manuscript

lower than chance level (measured d′ ≤ 0 in any condition). Data including these participants produced the same pattern as did the analyses report here. Materials—The same 240 words (60 in each condition) and their corresponding sentences in Experiment 3 were used again here. Each word was presented in either red or green font color and surrounded by a big (1000 × 280 pixels) or small (500 × 140 pixels) white box with a line width of 3 pixels. All words appeared in the same font size (36 points). Procedure—The same procedure as in Experiment 1 was adopted here. In short, each participant went through four study-test blocks in a unique order. Participants were asked to imagine the scene that the sentence depicts and rate the vividness of their mental image from 1 to 3 during the study phase. Their memory for the size and color of each word was later tested on a confidence scale from 1 to 6.

Author Manuscript

Analysis—The same statistical analyses as in Experiment 1 were performed again with data collected here. Results and Discussion Overall, performance (measured d′) on color and size recognition was significantly above chance in every condition, ts > 9.29, ps < 0.001. Average vividness was significantly lower in the Unitize Neither condition than the other three conditions, ps < .02. However, there was no significant difference between color accuracy and size accuracy of the same word in all four conditions, ts < 1.51, ps > .14 (Table 2). The DPSD model provided an excellent fit to the observed data in all four conditions with sum of squared errors ranging from 0.001– 0.003.

Author Manuscript Author Manuscript

Effect of encoding strategies on familiarity across conditions—Similar to Experiment 1, analysis on familiarity estimates for color showed a main effect of color encoding strategy, F(1,23) = 4.89, p = .04, ηp = .18, indicating a significantly higher contribution of familiarity to color recognition when color was unitized (in the Unitize Both and Unitize Color conditions) than when it was not (in the Unitize Size and Unitize Neither conditions). No main effect of size encoding strategy was found, F(1,23) = 3.47, p = .08, ηp = .13, but there was a significant interaction between encoding strategy and context type, F(1,23) = 7.67, p = .01, ηp = .25 (Figure 3A). Post hoc analyses showed that when color was unitized, familiarity estimates for color recognition were significantly greater when size was also unitized than when size was not unitized, Unitize Both vs. Unitize Color: t(23) = 3.45, p = .002, d = .70. However, when color was not unitized, familiarity estimates for color recognition were equally low regardless of whether size was or was not unitized, Unitize Size vs. Unitize Neither: t(23) = .60, p = .55, d = .12. The ROCs (see Figure 3B) of color recognition also showed that familiarity estimates were higher (more curvilinear) when unitization was applied to encode both color and size information (Unitize Both) than when it was only used on one detail (Unitize Color and Unitize Size) or never used for encoding (Unitize Neither). Consistent with Experiment 1, the ROC curve was similarly linear in the Unitize Color and Unitize Neither condition, reflecting a substantial influence of the non-unitized size detail on familiarity for the color J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 14

Author Manuscript

information. However, in Experiment 3 the Unitize Size condition (in which color was not unitized) actually produced a more curvilinear ROC than did the Unitize Color condition (in which color was unitized). Thus, it seems that the influence of a non-unitized, but extrinsic, size detail was even larger than that of the intrinsic size detail in Experiment 1.

Author Manuscript

Analysis on familiarity estimates for size showed a main effect of size encoding strategy. That is, familiarity estimates for size recognition were significantly higher when size was unitized (in the Unitize Both and Unitize Size conditions) than when size was not unitized (in the Unitize Color and Unitize Neither conditions), F(1,23) = 25.77, p < .001, ηp = .53. In contrast to Experiment 1, there was no main effect of color encoding strategy, F(1,23) = 1.55, p = .23, ηp = .06, nor significant interaction between encoding strategy and context type, F(1,23) = .12, p = .73, ηp = .01 (Figure 3C). Figure 3D shows that the ROCs of size recognition in the Unitize Both and Unitize Size conditions were both more curvelinear than those in the Unitize Color and Unitize Neither conditions, indicating that the effect of size unitization was not significantly affected by the encoding strategy used for color information.

Author Manuscript

Effect of encoding strategies on familiarity within conditions—In the Unitize Size condition, familiarity estimates for size (M = 1.38) were significantly higher than those for color (M = .80), t(23) = 3.37, p = .003, d = .69. This result indicates that the unitization of size information at encoding has a strong effect on familiarity-based size recognition, even in the presence of a relationally encoded intrinsic detail. Furthermore, although there was a trend showing that familiarity estimates for color (M = .86) were higher than those for size (M = .63) in the Unitize Color condition, the difference did not reach significance, t(23) = 1.48, p = .15, d = .30. On the other hand, familiarity estimates did not differ significantly between color recognition and size recognition in the Unitize Both condition, Mcolor = 1.41, Msize = 1.30, t(23) = .73, p = .47, d = .15, and in the Unitize Neither condition, Mcolor = .90, Msize = .80, t(23) = .70, p = .49, d = .14. Compared to the results from Experiment 1, current data showed a similar pattern of familiarity estimates in these two conditions when relational encoding and unitization were applied simultaneously in one episode, suggesting that the manner of visual presentation of the context detail does not explain the asymmetric unitization effect.

Author Manuscript

Effect of visual presentation on familiarity estimates—When data were collapsed across different conditions, familiarity estimates for intrinsic color (M = .99) were not significantly different from those for extrinsic size (M = 1.03), t(95) = .43, p = .67, d = .04. This differed from the results in Experiment 1, where intrinsic font size had a significantly greater contribution from familiarity at retrieval than did extrinsic font color. Thus, the increased familiarity for size information in Experiment 1 disappeared when size information was extrinsic rather than intrinsic. However, presenting color information as a visually intrinsic detail did not produce an increase in familiarity over size information. Across the two experiments, the familiarity estimates for size and color overall were greater in Experiment 3 than Experiment 1, color: MExp1 = .55, MExp3 = .99, t(190) = 4.70, p < .001, d = .68; size: MExp1 = .72, MExp3 = 1.03, t(190) = 3.16, p = .002, d = .46, regardless of whether the information was presented as perceptually intrinsic or extrinsic. J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 15

Author Manuscript Author Manuscript

Recollection estimates—Although we did not have predictions about the recollection estimates in this experiment, Table 2 summarizes recollection estimates in each condition. Except in the Unitize Size condition, in which recollection estimates for color (M = .40) were significantly higher than those for size (M = .26), t(23) = 3.60, p = .002, d = .73, there was no significant difference between recollection estimates for color and size judgments in all other conditions, ts < 1.43, ps > .16. Analysis on recollection estimates for color recognition showed that recollection estimates for color were higher when size was unitized than when it was not, F(1,23) = 38.09, p < .001, ηp = .62, but there was no main effect of color encoding strategy, F(1,23) = .83, p = .37, ηp = .04. There was also no significant interaction, F(1,23) = 1.59, p = .22, ηp = .07. In contrast, analysis on recollection estimates for size recognition showed that recollection estimates for size were significantly higher when color was unitized than when it was not, F(1,23) = 12.06, p = .002, ηp = .34, but there was no main effect of size encoding strategy, F(1,23) = 2.52, p = .13, ηp = .10. A significant interaction between context information and encoding strategy was found, F(1,23) = 4.84, p = .04, ηp = .17.

General Discussion

Author Manuscript

In the present study, we assessed the involvement of familiarity in source recognition when a unitized encoding strategy was applied alone or combined with relational association to encode multiple context details in one scenario. To our knowledge, these are the first experiments to investigate the effects of two encoding strategies independently applied to different context details. We found that there was no additive effect when unitization was used to encode more than one detail at the same time. That is, when unitization was the only encoding strategy used in an event, familiarity was involved in source recognition at a similar level whether one context detail was unitized (Unitize Color only, Experiment 2) or two context details were unitized (Unitized Both, Experiment 1). In addition, the binding of a relationally encoded detail to a unitized detail reduced the contribution of familiarity to retrieval of the unitized detail, which suggests that there is interference between the encoding of unitized and relational information.

Author Manuscript

Results from Experiments 1 and 3 showed that unitization of two independent context details produced a larger contribution from familiarity to recognition of either context detail individually. Comparison data from Experiment 2 further revealed that unitizing one detail (color) in the absence of a second manipulated detail produced similar levels of familiarity to unitizing both details. These findings support the notion that unitization of multiple context details would only increase the complexity or specificity of the new, unified, representation that was created, but that this increased complexity would not change the involvement of familiarity in recognition of any one context detail. In addition, we proposed that any context information could be relationally bound to a unitized representation in much the same way that it is bound to an item when not unitized. However, relational binding of contextual details adds additional cognitive demands to unitization of contextual details. When two encoding strategies were used to remember the details of one item, participants not only needed to form a new representation of the unitized information, but also had to encode the relational association between that new item representation and the non-unitized details at the same time. The binding of non-unitized context features may distract from J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 16

Author Manuscript

encoding of the unitized item representation and thus reduce the effectiveness of the unitization strategy during the encoding process.

Author Manuscript

A different explanation for these effects could be proposed based on the overlap, and therefore the interference, between the episodes in the unitized and non-unitized conditions. In the current paradigm, each non-unitized word was associated with an object that embodied the color or size property being assigned (stop sign/dollar bill or mouse/elephant), resulting in a common contextual feature (stop sign/dollar bill or mouse/elephant) shared among all non-unitized items in that condition. Although the color or size information did not produce explicit contextual overlap with the other unitized items in the same way as the non-unitized items, the “concept” of red/green or small/large was common to all unitized items. It is possible that increased interference across non-unitized contexts as compared to unitized contexts had some effect on our data. However, we would expect that increased interference would primarily affect the recollection factor, which drives retrieval of nonunitized contexts, rather than the familiarity factor. In fact, we found that recollection estimates were lower in the Unitize Neither conditions which may reflect the influence of interference on retrieval of non-unitized context details.

Author Manuscript

An alternative to the encoding-based explanation for the detrimental effects of combining unitization and relational binding on familiarity might involve changes in processing at retrieval. If retrieval of a relationally bound detail increases recollection of the unitized detail, there would likely be an accompanying decrease in the familiarity estimate for that unitized detail. However, the recollection estimates we collected do not support this explanation. In both Experiments 1 and 3, the recollection estimates for Unitize Color (color retrieval) and Unitize Size (size retrieval) are smaller than the recollection estimates for Unitize Both. Thus we must conclude that familiarity is being directly modulated by the combination of unitization and relational binding rather than indirectly modulated via effects on recollection. A more feasible retrieval-based explanation might be that retrieval of one unitized detail increases familiarity-based retrieval of the second detail, however our results from the Unitize Color Only condition suggest that the Unitize Both condition did not increase familiarity beyond the level of a single unitized detail but rather that the Unitize Size and Unitize Color conditions actually decreased familiarity from the level of a single detail. Thus it is not clear how a retrieval mechanism could explain the pattern of findings.

Author Manuscript

Our data suggest that relational encoding and unitization may not be processed independently in episodic memory, and the interaction seemed to depend to some degree on the context details that were encoded. Results from Experiments 1 and 3 suggest that the unitization effect on color was more easily compromised by the existence of non-unitized size information, compared to the unitization effect on size when there was also nonunitized color information in the scene. When unitized color was bound with relationally encoded size information (the Unitize Color condition), familiarity estimates for color recognition decreased, but when unitized size was bound with relationally encoded color information (the Unitize Size condition), familiarity estimates for size recognition remained high. We proposed three hypotheses for this discrepancy in size and color processing.

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 17

Author Manuscript

First, it is possible that some unitized sentences for color information were not as effective in attributing a new color to the item as others, which might make color unitization less likely to occur. However, analyses showed that our participants had equally vivid experiences with the sentences across conditions, and recognition accuracy was the same for color and size across the experiments. Therefore, any subtle difference in the nature of the encoding sentences does not explain why the effect of size unitization was stronger than that of color unitization. Second, Experiment 2 examined the hypothesis that the asymmetrical effect observed for color and size familiarity results from a difference in the nature of the associations made for non-unitized information in the Unitize Color and Unitize Size conditions. However, this hypothesis was not supported: whether size information was conveyed via an active item or via a passive item did not change familiarity estimates for color recognition.

Author Manuscript Author Manuscript

Third, Experiment 3 tested whether different visual cues manipulate the involvement of familiarity in addition to the effect of unitization. Previous studies have shown that intrinsic features are automatically processed together with the item, thus increasing familiarity in source recognition (Ecker et al., 2007a; Ecker, Zimmer, & Groh-Bordin, 2007b). For example, when associated with line drawings, color can be presented in a rectangular frame encasing the item in the extrinsic condition (e.g., a black line drawing of a hat in a green frame) or as a feature of the item itself in the intrinsic condition (e.g., a green silhouette of a hat). Event-related potentials (ERPs) recorded at retrieval showed that the early mid-frontal ERPs associated with familiarity (FN400 component) significantly decreased when the color changed from study to test in the intrinsic condition but not in the extrinsic condition. Moreover, the reaction time for hit responses was significantly longer in the extrinsic condition than in the intrinsic condition, suggesting that retrieving an extrinsic feature may be less automatic (Ecker et al., 2007b). However, we found asymmetric patterns of familiarity contribution both when size was intrinsic while color was extrinsic (Experiment 1) and when size was extrinsic while color was intrinsic (Experiment 3). Therefore the manner of perceptual presentation does not explain the differences in unitization of size and color information.

Author Manuscript

Given the consistent asymmetry between familiarity for color and size details was not caused by differential vividness of the sentences, active vs. passive processing, extrinsic vs. intrinsic presentation, or differences in recognition accuracy, we propose that the nonreciprocal unitization effect for color and size may reside in the nature of each context detail. Different types of context details may have properties that facilitate or limit accessibility for unitization. For example, the size of an item is not as readily interchangeable as the color of an item. In addition, in the Unitize Size condition, participants were strongly encouraged to form a mental image that unitized size information with the item and then associate that unitized representation to a stop sign or a dollar bill. Because size is typically perceived in relation to another object, the stop sign or dollar bill in the sentence may have served as a reference to compare with the size of the item and become part of the reason why the item is big or small. Therefore, color information was tagged along with the unitized scene instead of disrupting it, generating minimal to no detrimental effect of non-unitized color information on size unitization. However, in the

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 18

Author Manuscript

Unitize Color condition, when participants formed a mental image with unitized color information, the elephant or mouse that we provided to associate with that unitized representation may have become the reference for size perception. Because color can be perceived by viewing just the item alone, making relational associations not only had no contribution to color perception but also might distract our participants’ attention from unitizing the color information, thus creating a detrimental effect of non-unitized size information on color unitization. Therefore, the unitization process may be partially limited or facilitated by the nature of the information being unitized. This finding highlights the importance of comparing unitization to relational encoding within a single class of context details in order to avoid differences due solely to the nature of the context detail itself.

Author Manuscript

Finally, relational memory theory (Cohen & Eichenbaum, 1993) and the Binding of Item and Context theory (Diana, Yonelinas, & Ranganath, 2007) both propose that the hippocampus creates a single, unified representation for an event. Therefore, the unitized context and item information would be bound with the remaining, separately encoded, context information via the hippocampal binding process. These theories suggest that the unitized item/context should interact during encoding processing with any additional context details that are relationally bound to the episode. This hypothesis is supported by the current results. The effects of unitization and relational binding do not appear to be independent of one another. However, we did not find evidence for a more extreme interpretation of these theories: that unitization of some context details increases familiarity for relationally-bound context details.

Author Manuscript Author Manuscript

In conclusion, we found that when unitization is the sole encoding strategy used in an episode, there is no additive effect related to how many context details are unitized with the same item. If we return to the wedding event example from the introduction, this means that unitizing Pam, Pat, and Pat’s child into a single representation does not appear to increase the contribution of familiarity to retrieving an association with any one of those item details as compared to unitizing Pam and Pat alone. In addition, unitization and relational encoding interact with each other when applied simultaneously. Depending on the context detail being encoded, the presence of a non-unitized context detail may impose a detrimental effect on the contribution of familiarity to retrieval of another unitized context detail. Combining unitization with relational encoding appears to require more cognitive resources than encoding the same information through unitization alone. That is, encoding specific details, for example the music being played during the wedding event, may affect the creation of the new “Pam and Pat” item unit such that it is less likely that familiarity will enable retrieval of the two as a group in future retrieval events. These results suggest that unitization operates as a single, cohesive, process at encoding, and therefore unitizing two context details with an item is no more difficult than unitizing a single context detail. Future investigations into the corresponding brain activity when using different combinations of encoding strategies may be informative to understand how these strategies function to provide a complete episodic memory.

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 19

Author Manuscript

Acknowledgments This work was supported by National Institute of Mental Health Grant R00MH083945. We thank Roger W. Strong for assistance with data collection and Tanner M. Hurley for editing the manuscript.

References

Author Manuscript Author Manuscript Author Manuscript

Boywitt CD, Meiser T. Bound context features are integrated at encoding. The Quarterly Journal of Experimental Psychology. 2012a; 65(8):1484–1501. [PubMed: 22493990] Boywitt CD, Meiser T. The role of attention for context-context binding of intrinsic and extrinsic features. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2012b; 38(4): 1099–1107. Cohen, NJ.; Eichenbaum, H. Memory, Amnesia, and the Hippocampal System. Cambridge, MA: The MIT Press; 1993. Davachi L. Item, context and relational episodic encoding in humans. Current Opinion in Neurobiology. 2006; 16(6):693–700. [PubMed: 17097284] Diana RA, Yonelinas AP, Ranganath C. Imaging recollection and familiarity in the medial temporal lobe: a three-component model. Trends in Cognitive Sciences. 2007; 11:379–386. [PubMed: 17707683] Diana RA, Yonelinas AP, Ranganath C. The effects of unitization on familiarity-based source memory: Testing a behavioral prediction derived from neuroimaging data. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2008; 34(4):730–740. Ecker UKH, Zimmer HD, Groh-Bordin C. Color and context: An ERP study on intrinsic and extrinsic feature binding in episodic memory. Memory & Cognition. 2007a; 35(6):1483–1501. [PubMed: 18035643] Ecker UKH, Zimmer HD, Groh-Bordin C. The influence of object and background color manipulations on the electrophysiological indices of recognition memory. Brain Research. 2007b; 1185:221–230. [PubMed: 17950711] Giovanello KS, Keane MM, Verfaellie M. The contribution of familiarity to associative memory in amnesia. Neuropsychologia. 2006; 44(10):1859–1865. [PubMed: 16643967] Graf P, Schacter DL. Unitization and grouping mediate dissociations in memory for new associations. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1989; 15(5):930–940. Haskins AL, Yonelinas AP, Quamme JR, Ranganath C. Perirhinal cortex supports encoding and familiarity-based recognition of novel associations. Neuron. 2008; 59(4):554–560. [PubMed: 18760692] Hockley WE, Consoli A. Familiarity and recollection in item and associative recognition. Memory & Cognition. 1999; 27(4):657–664. [PubMed: 10479824] Macken WJ. Environmental context and recognition: The role of recollection and familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2002; 28(1):153–161. Mandler G. Recognizing: The judgment of previous occurrence. Psychological Review. 1980; 87(3): 252–271. Meiser T, Bröder A. Memory for multidimensional source information. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2002; 28(1):116–137. Mulligan NW. Generation disrupts memory for intrinsic context but not extrinsic context. The Quarterly Journal of Experimental Psychology. 2011; 64(8):1543–1562. [PubMed: 21500106] Parks CM, Yonelinas AP. Moving beyond pure signal-detection models: Comment on Wixted. Psychological Review. 2007; 114(1):188–202. [PubMed: 17227187] Quamme JR, Yonelinas AP, Norman KA. Effect of unitization on associative recognition in amnesia. Hippocampus. 2007; 17(3):192–200. [PubMed: 17203466] Ranganath C, Yonelinas AP, Cohen MX, Dy CJ, Tom SM, D’Esposito M. Dissociable correlates of recollection and familiarity within the medial temporal lobes. Neuropsychologia. 2004; 42(1):2– 13. [PubMed: 14615072]

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 20

Author Manuscript Author Manuscript Author Manuscript

Rugg MD, Vilberg KL. Brain networks underlying episodic memory retrieval. Current Opinion in Neurobiology. 2013; 23(2):255–260. [PubMed: 23206590] Rugg MD, Yonelinas AP. Human recognition memory: a cognitive neuroscience perspective. Trends in Cognitive Sciences. 2003; 7(7):313–319. [PubMed: 12860190] Staresina BP, Davachi L. Differential encoding mechanisms for subsequent associative recognition and free recall. Journal of Neuroscience. 2006; 26(36):9162–9172. [PubMed: 16957073] Turriziani P, Fadda L, Caltagirone C, Carlesimo GA. Recognition memory for single items and for associations in amnesic patients. Neuropsychologia. 2004; 42(4):426–433. [PubMed: 14728917] Uncapher MR, Otten LJ, Rugg MD. Episodic encoding is more than the sum of its parts: An fMRI investigation of multifeatureal contextual encoding. Neuron. 2006; 52(3):547–556. [PubMed: 17088219] Watson HC, Wilding EL, Graham KS. A role for perirhinal cortex in memory for novel object-context associations. Journal of Neuroscience. 2012; 32(13):4473–4481. [PubMed: 22457495] Weis S, Specht K, Klaver P, Tendolkar I, Willmes K, Ruhlmann J, Fernández G. Process dissociation between contextual retrieval and item recognition. Neuro Report. 2004; 15(18):2729–2733. Wixted JT. Dual-process theory and signal-detection theory of recognition memory. Psychological Review. 2007; 114(1):152–176. [PubMed: 17227185] Yonelinas AP. Recognition memory ROCs for item and associative information: The contribution of recollection and familiarity. Memory & Cognition. 1997; 25(6):747–763. [PubMed: 9421560] Yonelinas AP. The contribution of recollection and familiarity to recognition and source-memory judgments: A formal dual-process model and an analysis of receiver operating characteristics. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1999; 25(6):1415–1434. Yonelinas AP. Components of episodic memory: the contribution of recollection and familiarity. Philosophical Transactions of the Royal Society B: Biological Sciences. 2001; 356(1413):1363– 1374. Yonelinas AP. The nature of recollection and familiarity: A review of 30 years of research. Journal of Memory and Language. 2002; 46(3):441–517. Yonelinas AP, Kroll NEA, Dobbins IG, Soltani M. Recognition memory for faces: When familiarity supports associative recognition judgments. Psychonomic Bulletin & Review. 1999; 6(4):654–661. [PubMed: 10682209] Yonelinas AP, Parks CM. Receiver operating characteristics (ROCs) in recognition memory: A review. Psychological Bulletin. 2007; 133(5):800–832. [PubMed: 17723031] Zimmer HD, Ecker UKH. Remembering perceptual features unequally bound in object and episodic tokens: Neural mechanisms and their electrophysiological correlates. Neuroscience & Biobehavioral Reviews. 2010; 34(7):1066–1079. [PubMed: 20138910]

Author Manuscript J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 21

Author Manuscript Author Manuscript Figure 1.

Author Manuscript

Familiarity estimates of color recognition (A) and size recognition (C), based on the DPSD model fits to the observed data, in all four conditions in Experiment 1. Familiarity estimate of color recognition in the Color Only condition in Experiment 2 is also included in A for comparison. The ROCs of color recognition (B) and size recognition (D) in probability space. The ROC for the Color Only condition in Experiment 2 is also included in D for comparison.

Author Manuscript J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 22

Author Manuscript Author Manuscript

Figure 2.

Familiarity estimates of color recognition (A) and size recognition (B) in the Active Size (solid line) and Passive Size (dashed line) conditions in Experiment 2. No significant difference was found between familiarity estimates in these conditions.

Author Manuscript Author Manuscript J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 23

Author Manuscript Author Manuscript Author Manuscript

Figure 3.

Familiarity estimates of color recognition (A) and size recognition (C), based on the DPSD model fits to the observed data, in all four conditions in Experiment 3. The ROCs of color recognition (B) and size recognition (D) in probability space.

Author Manuscript J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Tu and Diana

Page 24

Table 1

Author Manuscript

Examples of words and sentences used in the four conditions: Unitize Color, Unitize Size, Unitize Both, and Unitize Neither. Condition

Color information

Size information

Unitize Color

Unitized

Not unitized

Example sentence The WOUND is red because it has not yet healed from the mouse bite.

Unitize Size

Not unitized

Unitized

The COOKIE is small because the buyer can only afford one dollar bill.

Unitize Both

Unitized

Unitized

The BOWL is big and green because the chef just made basil pesto for a party of 200 people.

Unitize Neither

Not unitized

Not unitized

The SNOWMAN beside the stop sign was knocked down by an elephant.

Author Manuscript Author Manuscript Author Manuscript J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01.

Author Manuscript

Author Manuscript

Author Manuscript

2.23

2.30

2.19

Unitize Size

Unitize Both

Unitize Neither

2.31

2.29

Active Size

Passive Size

2.25

2.22

2.28

2.11

Unitize Color

Unitize Size

Unitize Both

Unitize Neither

Experiment 3

2.37

Color Only

Experiment 2

2.18

Unitize Color

Experiment 1

Vividness

J Exp Psychol Learn Mem Cogn. Author manuscript; available in PMC 2017 January 01. .76

.78

Size

.89

Size

.87

Color

.87

Size

.82

Color

.28

.74

Size Color

.28

.25

.20

.13

.22

.21

.22

.30

.22

.27

.24

NA

.15

.34

.28

.26

.17

.29

.32

.28

.29

FA rate

.74

Color

.74

.66

Size

.77

Size

.72

Color

NA

Size

.81

Color

Color

.66

.66

Color Size

.83

.78

Color Size

.73

.68

Color Size

.66

.68

Size

Color

Hit rate

1.41

1.55

2.28

2.44

2.05

1.99

1.32

1.51

1.22

1.25

1.47

1.36

NA

2.04

.89

1.05

1.78

1.88

1.25

1.00

1.07

1.10

Measured d′

.80

.90

1.30

1.41

1.38

.80

.63

.86

.73

.65

.72

.52

NA

.88

.47

.55

1.08

.83

.76

.39

.57

.44

Familiarity estimates

.27

.25

.39

.47

.26

.40

.30

.25

.19

.23

.30

.31

NA

.41

.17

.19

.32

.39

.21

.25

.20

.26

Recollection estimates

Average vividness ratings, hit rates, false alarm rates, measured d′, familiarity estimates and recollection estimates from the DPSD model fits for color and size information of each condition in Experiment 1, 2, and 3.

Author Manuscript

Table 2 Tu and Diana Page 25

Two are not better than one: Combining unitization and relational encoding strategies.

In recognition memory, recollection is defined as retrieval of the context associated with an event, whereas familiarity is defined as retrieval based...
NAN Sizes 0 Downloads 9 Views