Forgetting “Novel” but Not “Dragon”: The Role of Age of Acquisition on Intentional and Incidental Forgetting

Alejandra Marful; Carlos J. Gómez-Ariza; Analía Barbón; Teresa Bajo

doi:10.1371/journal.pone.0155110

Abstract

Two experiments studied how the age at which words are acquired (Age of Acquisition, AoA) modulates forgetting. Experiment 1 employed the retrieval-practice paradigm to test the effect of AoA on the incidental forgetting that emerges after solving competition during retrieval (i.e., retrieval-induced forgetting, RIF). Standard RIF appeared with late-acquired words, but this effect disappeared with early-acquired words. Experiment 2 evaluated the effect of AoA on intentional forgetting by employing the list-method directed forgetting paradigm. Results showed a standard directed forgetting effect only when the to-be-forgotten words were late-acquired words. These findings point to the prominent role of AoA in forgetting processes.

Citation: Marful A, Gómez-Ariza CJ, Barbón A, Bajo T (2016) Forgetting “Novel” but Not “Dragon”: The Role of Age of Acquisition on Intentional and Incidental Forgetting. PLoS ONE 11(5): e0155110. https://doi.org/10.1371/journal.pone.0155110

Editor: Philip Allen, University of Akron, UNITED STATES

Received: November 13, 2015; Accepted: April 25, 2016; Published: May 10, 2016

Copyright: © 2016 Marful et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper and its Supporting Information files.

Funding: This research was supported by grants from the Spanish Ministry of Economy and Competitiveness (http://www.mineco.gob.es/portal/site/mineco/?lang_choosen=en) to AM (PSI2013-46033-P), TB (PSI2012-33625), and CJG-A (PSI2011-25797), and from the Ministry of Science (http://www.mineco.gob.es/portal/site/mineco/?lang_choosen=en) (EDU2008-01111) and the Andalusian Government (http://www.juntadeandalucia.es/) (P12-CTS-2369) to TB. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Skills and concepts learned during early childhood seem to have a special status in cognition. Thus, the ease with which one specific representation is processed can be determined by the age at which it was acquired (Age of Acquisition, AoA). For example, certain musical abilities, such as absolute pitch, can usually only be learnt during childhood [1], and concepts that are acquired early in life are processed faster than late-acquired ones [2–3].

An advantage for early-acquired words has been reported in many psycholinguistic domains such as picture naming, word reading, and auditory and visual lexical decision tasks [3–6]. In addition, early-acquired words have been shown to be more resistant to cognitive impairment in Alzheimer’s disease and other neuropsychological problems [7–8]. This advantage for early-acquired representations has been attributed, for example, to better connections in semantic memory [9,10]. It has also been suggested that early-acquired words constitute the basis by which the mental network is organized. Accordingly, once a mental configuration is created with these first representations, it losses plasticity so that later acquired representations are more difficult to integrate ([11], see also [12–14] for a review and for alternative hypotheses).

Despite its relevance for cognition, most of the studies on AoA have primarily focused on the language domain. The link between AoA and other cognitive functions such as episodic memory has been much less investigated, with available studies providing mixed results [15–18]. One intriguing question in memory research is why we can easily forget some information while other memories seem to be indelible. Given the special status of early-acquired knowledge representations, we aimed to explore whether AoA plays a role in forgetting. Thus, we tested the effect of AoA on two types of forgetting phenomena that are thought to be aftereffects of exerting control over unwanted or irrelevant memories [19]. In Experiment 1, we focused on retrieval-induced forgetting (RIF), an incidental type of forgetting whereby practicing retrieval (retrieval-practice, RP) of some studied items causes the temporary inaccessibility of related non-target items that compete for retrieval. From an influential standpoint [20, 21], competing memories during retrieval are subjected to inhibitory control which helps facilitate the retrieval of target memories. Hence, if once non-target traces turn out to be relevant later, as is the case in the RP procedure, they become less recallable because of the residual effect of inhibition. In Experiment 2, we focused on list-method directed forgetting (LM-DF), an intentional type of forgetting that is normally observed after providing participants with the instruction to forget previously encoded items. According to a well-established account (for an alternative account, see [22]), directed forgetting is caused by a transitory state of inhibition of the to-be-forgotten items which is triggered to prevent these items from hindering learning/recall of new items [23–24]. While differing in a variety of aspects, RIF and LM-DF are thought to reflect the adaptive value of making unwanted information less accessible.

On the basis of the previously described cognitive advantages of early-acquired words, we expected early-acquired items to be more resistant to both intentional and incidental forgetting than later-acquired ones. Our hypothesis is that because early acquired words seem to be better interconnected and integrated in memory, they should be less susceptible to both incidental and intentional forgetting effects.

Experiment 1

RIF has been widely studied using the retrieval-practice procedure [21]. In this paradigm, category-exemplar pairs are presented for study (fruit-orange, fruit-apple, etc.). Then, during a retrieval-practice phase, participants are cued to repeatedly recall half of the items from half the categories (practiced items, Rp+; fruit ap____). Finally, participants’ memory of all the studied items is tested. Results typically show that unpracticed items from practiced categories (Rp- items; orange) tend to be more poorly recalled than control (Nrp) items (RIF effect). Whereas different mechanisms have been posited to account for RIF (see [25] for a context-change account), memory inhibition has emerged as a potential candidate [21, 26–28]. From this viewpoint, the memory impairment for competitor (Rp-) items is interpreted as the aftereffect of their inhibition during retrieval practice.

Starting from the premise that early-acquired words hold a special status in cognition, and, regardless of the mechanism underlying RIF, one would expect these words, relative to late-acquired ones, to be less prone to incidental forgetting. This expectation might seem at odd with the frequency effect reported by Anderson, Bjork, and Bjork (1994) [29]. In their experiments they manipulated the category frequency of the competitor exemplars and reported larger RIF effects for high than low frequency words, suggesting that highly accessible representations are more vulnerable to retrieval-induced forgetting. Note, however, that category frequency in Anderson et al’s study refers to the strength of the category-exemplar connections, with high frequency exemplars being associated with higher probabilities of being activated upon the category presentation. This would in turn produce stronger competition relative to weaker elements with proportional suppression acting upon these competing representations. In contrast, although early representation may also be more accessible, we assume that they are more resistant to forgetting because their internal structure (more interconnected or better consolidated) protect them from competition and forgetting. Supporting this hypothesis, previous studies have shown that conditions fostering integration through instructions [30], or by the nature of the materials [31–33] reduce forgetting effects.

Methods

Participants.

Forty Spanish students (36 women; M = 20.07, SD = 2.70) from the universities of Granada and Jaén participated in this experiment in exchange for course credit. Sample size was determined on the basis of prior studies [34]. The study was approved by the local ethics committees (Universities of Jaén and Granada). Participants gave written informed consent.

Materials.

Eight orthographically based categories (i.e., semantically unrelated words that started with the same two letters: CA-camino (path), CA-cachorro (puppy), etc.; see [34]) each including six words were created from Davies et al. [4] database. Half of them were composed of early-acquired words (i.e., 24 words), whereas the other half comprised late-acquired words (i.e., 24 words).

Late- and early-acquired lists (see S1 Table) differed in age of acquisition scores [4] F(1,46) = 134.67, p < .0001, η² = .74 but were matched in word frequency [35] from the NIPE database [36] F(1,46) = .36, p = .55, η² 008, number of phonological F(1,46) = 2.287, p = .14, η² .05 and orthographic F(1,46) = 3.67, p = .06, η² = .07 neighbors, number of words beginning with the first two letters F(1,46) = 2.085, p = .16, η² = .04 and number of letters F(1,46) = .009, p = .92, η² = .00 but differed in concreteness F(1,46) = 15.3, p <. 0001, η² = .25 [37]. Semantic and associative relationships among words were kept to a minimum; there was no phonological overlap between the first two letters of the words from the different categories, and all words were between two and five syllables. Two more six-words sets were created to be used as fillers.

For the final memory test, two or three letters (depending on the word length) of each word were removed, with one of these always being the first letter. We did this to minimize retrieval interference from Rp+ over Rp- items [21].

Procedure.

Participants were instructed to study a list of unrelated words sharing the first two letters for a later memory test. To avoid floor effect due to very long lists, early-acquired lists were presented to half of the participants and late-acquired lists to the remaining students. Each item was presented individually next to its first two letters (e.g., CA-camino) in the center of the screen (5000 ms), preceded by a blank screen (1000 ms) and a fixation point (1200 ms) and followed by a blank screen (500 ms). The complete list of stimuli was presented twice for study in a random order. Six filler items appeared at the beginning and 6 at the end of the study phase to minimize primacy and recency effects. During the retrieval-practice phase participants performed repeated retrieval of half of the items from half of the categories (12 Rp+ items). The practiced items and, consequently, the items that served as Rp- and Nrp- varied across two counterbalance conditions. Following a pre-cueing procedure [34], a fixation point appeared (1200 ms), followed by the lexical category cue (CA) (2000 ms), a blank screen (1000 ms), and the word stem (car) for retrieval (6000 ms). Each Rp+ item appeared four times.

After a five-minute distractor task participants performed an unexpected final memory test where fragments of the previously studied words were presented for retrieval (-am-no to retrieve camino). Each fragment (8000 ms) was preceded by a fixation point (1200 ms) and followed by a blank screen (1000 ms). To minimize output interference, fragments corresponding to the unpracticed (Rp- and Nrp-) items appeared before those fragments corresponding to the practiced items.

Design.

The experiment used a 2 (item type: Rp-, Nrp-; within-participants) X 2 (age of acquisition; early-, late-acquired words; between-participants) mixed-factorial design. Notice that since two counterbalanced conditions were created, Rp- items in one condition were compared to the same items that served as baseline (Nrp-) items in the other condition (the same type of correspondence occurred for Rp+ and Nrp+ items).

Results

Retrieval practice.

Average recall success did not differ between conditions (late condition, M = .56, SD = .22; early condition, M = .46, SD = .22), F(1,38) = 2.14, p = .15, η² = .05).

Final test.

A 2 (item type: Rp-, Nrp-) x 2 (AoA: early, late) repeated measures ANOVA showed that the main effect of item type was not significant, F(1,38) = 1.16, p = .29, = .03. See Table 1. More importantly, the main effect of AoA F(1,38) = 13.85, p = .001, = .27 was modulated by the item type x AoA interaction F(1,38) = 4.86, p = .03, = .12. Thus, while late-acquired words showed the standard RIF effect (competitor, Rp-, items were more poorly recalled than baseline, Nrp-, items, F(1,19) = 5.48, p = .03, η² = .22, RIF was not present in the early-acquired words condition F(1,19)<1, η² = .03.

Download:

Table 1. Final memory test recall (standard deviations in brackets) in Experiment 1.

https://doi.org/10.1371/journal.pone.0155110.t001

Although not relevant for our hypothesis, for completeness we also checked for facilitation effects by means of a 2 (item type: Rp+, Nrp+) x 2 (AoA: early, late) repeated-measures ANOVA (see Table 1 and S2 Table). Neither the main effects nor the interaction (all with F < 1) reached statistical significance. This lack of overall facilitation is consistent with previous studies that have also employed independent cues (i.e., memory cues different than the ones used to do retrieval practice) during the final memory test, and suggest that unlike forgetting, the benefit of retrieval practice may depend on cue-overlap between the practice phase and the memory test [38–39]. Note also that to minimize their influence over Rp- items’ recall, the cues for Rp+ items appeared at the end of the memory test and this could also play a role in reducing facilitation.

Discussion

Experiment 1 revealed that while the standard RIF effect was observed with late-acquired words, there was no forgetting effect with early-acquired ones. Hence, the early acquisition of a word seems to protect it from incidental forgetting, at least during repeated retrieval of related items. This finding suggests that early-acquired representations, which are assumed to be strongly consolidated and overlearned, play a prominent role in cognition possibly because they are better interconnected and more efficiently represented in the memory system (see, for example, [12–14] for a review of the AoA explanatory hypothesis).

As mentioned, although at first glance the absence of RIF for early-acquired words might seem at odd with previous literature showing larger RIF effects for high-frequency words [29], it is consistent with studies reporting that RIF only emerges if Rp- items compete with Rp+ items during retrieval practice and it is reduced when items integration is promoted (e.g., [34,40]). In this line, previous studies have shown that the memory impairment for Rp- items decreases, disappears or is less durable in experimental settings when the nature of the to-be-studied materials induces the building of associative or semantic links [31–33], or when participants (spontaneously or following the experimenter’s instructions) establish these connections [30]. Hence, we posit that, unlike high-frequency words, early-acquired words, which are highly interconnected and better represented in the memory system, are less susceptible to inhibition and forgetting during retrieval practice.

The absence of RIF in the early AoA condition does not seem to be related to a deflated baseline condition. In fact, early-acquired Nrp- items were better recalled than late-acquired Nrp- items, with performance in both control conditions within the range of previous RIF studies [41]. However, it could be argued that the observed differences between early and late acquired words regarding RIF could be explained in terms of resources availability. This argument would suggest that retrieval of early-acquired words could involve less cognitive resources (less cognitive load) than late- acquired ones and, as a consequence, no forgetting would show up. Contrary to this argument, however, previous findings indicate that it is precisely the presence of high-cognitive load conditions what may prevent RIF from emerging [42–43]. Because selective retrieval is thought to place demands on attentional control to overcome interference from competing memories [20], less cognitive resources available during retrieval practice would make inhibition harder to implement. In our experiment, however, it was the late AoA condition (presumably the one demanding the highest cognitive load) that exhibited the aftereffect of inhibitory control (RIF).

Experiment 2

In order to explore whether the AoA effect generalizes to other situations requiring memory control, in Experiment 2 we examined the role of AoA in intentional forgetting by means of the LM-DF paradigm [23]. A first list of items (L1) was presented for study followed by a cue to either recall (remember group) or forget (forget group) the studied list. Participants then studied a second list (L2) and, finally, memory of List 1 was tested. The forget group typically exhibits a memory impairment for List 1 relative to the remember group. This directed-forgetting effect (DF) has been attributed to a transitory state of inhibition of the to-be-forgotten list triggered by the instruction to forget [19, 23–24]; for a non-inhibitory account, see [22]. To assess whether AoA may also modulate the DF, a list of early- or late-acquired words was presented for study followed by either recall or forget instructions. To the extent that AoA may diminish intentional forgetting, we expected larger DF effects with late-acquired than early-acquired words.