Skip to main content
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2024 Oct 16;18(4):e12753. doi: 10.1111/tops.12753

A Matter of Memory? Age‐Invariant Relative Clause Disambiguation and Memory Interference in Older Adults

Willem S van Boxtel 1,✉, Laurel A Lawyer 2
PMCID: PMC13740940  PMID: 39415427

Abstract

Past research suggests that Working Memory plays a role in determining relative clause attachment bias. Disambiguation preferences may further depend on Processing Speed and explicit memory demands in linguistic tasks. Given that Working Memory and Processing Speed decline with age, older adults offer a way of investigating the factors underlying disambiguation preferences. Additionally, older adults might be subject to more severe similarity‐based memory interference given their larger vocabularies and slower lexical access. Nevertheless, memory interference and sentence disambiguation have not been combined in studies on older adults before. We used a self‐paced reading paradigm under memory load interference conditions. Older (n = 30) and Younger (n = 35) readers took part in the study online; reading times were recorded and measures of comprehension accuracy and load recall were collected. This setup allowed for the implicit measurement of attachment biases and memory interference effects interactively. Results show that similarity‐based interference affected both age groups equally, but was more pronounced in NP2‐biased structures, which took participants generally longer to read. Attachment preferences did not differ by group and were unaffected by Working Memory span. However, accuracy on recall prompts was predicted by Working Memory span in both groups. Findings of greater interference in syntactically dispreferred structures support unified processing models where parsing constraints naturally interact. The lack of age differences on our measures further aligns with research finding age‐invariant implicit language processing.

Keywords: Sentence processing, Aging, Relative clause disambiguation, Memory interference, Working memory, Processing speed

Short abstract

Disambiguation preferences are a crucial metric for examining language across the adult lifespan. Past research has tied disambiguation to working memory capacity, which declines in adulthood. This study took a novel approach to investigate disambiguation and working memory by examining similarity‐based interference, and found no age differences between younger and older adults.

1. Introduction

Relative clause disambiguation tasks have been at the forefront of psycholinguistic research for decades and continue to be a relevant paradigm to the field as reliable indicators of linguistic parsing strategies (e.g., Carreiras & Clifton Jr, 1993; Checa‐Garcia, 2016; Frazier & Rayner, 1982; Mak, Vonk, & Schriefers, 2002). However, little is known about whether disambiguation preferences shift with greater age (cf. Payne et al., 2014). In sentences such as (1), it is ambiguous whether the referent of the relative clause is “the owner” (NP1) or “the house” (NP2). Sentences may also be semantically disambiguated, as in (2) and (3), to an NP1 or NP2 interpretation. Native speakers of English have generally been found to favor a strategy of attaching the relative clause to the most recently processed referent—in this case, NP2 (known as Recency; Altmann, van Nice, Garnham, & Henstra, 1998; Gibson et al., 1996).

  • 1.

    The owner of the house that looked old appealed for help.

  • 2.

    The owner of the house that had the moustache appealed for help.

  • 3.

    The owner of the house that needed renovation appealed for help.

Some evidence suggests these attachment disambiguation preferences shift with advancing age due to declining Working Memory (WM) capacity (Kemtes & Kemper, 1997; Payne et al., 2014). Specifically, Felser, Marinis, and Clahsen (2003) suggested syntactic chunking of sentences by readers with lower WM spans leads to different attachment preferences. Felser et al. claim readers store NP1 in WM, then insert a prosodic break, and only then move on to storing NP2. The first NP may, therefore, be retained in WM more prominently, allowing for a more stable integration into the WM system and leading to a greater likelihood of its retrieval (see further Abney, 1991; Beese et al., 2017; Frazier & Fodor, 1978).

However, the Felser et al. (2003) account was challenged by Swets, Desmet, Hambrick, and Ferreira (2007) and Traxler (2009), who encouraged participants to insert a reading break before and after a relative clause. For instance, the sentence in (1) would be segmented on‐screen as “The owner of the house // that looked old // appealed for help.” While low‐span readers were found to prefer NP1 when sentences were unsegmented, high‐span readers preferred NP1 attachment only when sentences were presented in chunks. According to Swets et al. (2007), inserting chunks resulted in all participants reading as low‐span individuals do.

Following this, older adults should show a preference for NP1, or a less defined bias toward NP2. This theory was tested in a reading study by Payne et al. (2014), who discovered a general preference for NP2, but the smaller a participant's WM span, the less likely they were to show this preference. Older adults showed significantly stronger effects of this modulation than younger adults. Notably, Payne et al. (2014) report varying attachment preferences based on the measure recorded: while on‐line reading times suggested a general NP2 bias, results from off‐line, post‐hoc comprehension questions were less clear cut. This suggests differences between off‐line and on‐line, explicit and implicit measures when used with older and younger adults (see further James, Fraundorf, Lee, & Watson, 2018; Waters & Caplan, 1996). Indeed, syntactic processing measures which require conscious (or explicit) memory recall are more challenging for older readers than unconscious (or implicit) processes (Bopp & Verhaeghen, 2005; Kemper & Herman, 2006; Yang, Chen, He, & Merrill, 2020). However, it is still unclear to what extent different memory types are involved in the processing of syntactic ambiguities.

Additionally, the hypothesis that lower‐span readers are more biased toward NP1 compared to higher‐span individuals also appears to contrast with abundant evidence around decay of information stored in WM (Lemaire & Portrat, 2018; Werner, 2019). Upon storage of informational features in memory, these memory chunks retain a set amount of activation until retrieval operations recall the information at the right moment. If too much material intervenes between the points of storage and retrieval, activation of the memory chunks decays to the point where retrieval becomes more difficult (Lewis, Vasishth, & Van Dyke, 2006; Lemaire, Pageot, Plancher, & Portrat, 2018). Following this notion, in relative clause sentences, the activation of NP1 would have decayed further than the activation of NP2 when the relative clause needs resolving. Nevertheless, Swets et al. (2007), Payne et al. (2014), and Traxler (2009) all suggest NP1 is the more likely referent for readers with lower WM capacity. This objection is especially relevant to older adults, as age‐related changes in Processing Speed are likely to affect activation decay (Salthouse, 1996; Salthouse & Coon, 1993; Salthouse & Babcock, 1991; Salthouse, 2000).

We nevertheless make a theoretical distinction between the constructs of WM and Processing Speed. While the two measures are frequently co‐concurrent in older adults (e.g., Salthouse, 1996), they functionally target different cognitive skills. While WM is concerned with the active maintenance of information in the memory system while processing is ongoing, Processing Speed refers to more domain‐general skills which encompass a wide range of cognitive functions. Findings from clinical populations also support a functional distinction between WM and Processing Speed: for instance, patients with acquired neurogenic language disorders consistently present with WM impairments (e.g., Mayer & Murray, 2012; Martin, Kohen, & Kalinyak‐Fliszar, 2010), yet observed delays in Processing Speed are inconsistent and exhibit wide‐ranging variability (Yoo, Dickey, Yoo, & Dickey, 2022). We consider these findings as evidence for a functional distinction between the mechanisms supporting WM and Processing Speed.

Returning to the current paradigm, very few studies have investigated whether the faster decay of cues stored in WM would bias older adults to NP2 (cf. Payne et al., 2014). However, addressing memory capacity and decay is not the only angle from which to consider WM. Memory effects on syntax processing have also been investigated based on the quality rather than the quantity of WM. In particular, the finding that retrieval may be subject to semantic interference has received considerable attention (e.g., Gordon, Hendrick, & Johnson, 2001; Gordon, Hendrick, & Levine, 2002; Gordon, Hendrick, Johnson, & Lee, 2006; Lewis, Vasishth, & Van Dyke, 2006; Lemaire & Portrat, 2018; Oberauer & Lin, 2017). For instance, Van Dyke and McElree (2006) presented participants with three memory load words before they paced through sentences. Memory loads were either semantically similar or different to the semantic content in the trial sentence. Similar loads were found to result in significant reading slowdowns (see further Gordon et al., 2006). It is unknown whether older adults' declining scores on WM measures are associated with increased susceptibility to similarity‐based interference.

Combinations of this framework with research on language and aging are virtually nonexistent. One exception is a study by Kemper and Herman (2006), which showed that younger adults are affected by memory interference while reading sentences even though older adults were not. In Kemper and Herman's study, memory interference was manipulated in a similar manner to Van Dyke and McElree (2006), by presenting semantically matching or mismatching load words before asking participants to read sentences. In addition, Kemper and Herman (2006) manipulated the syntactic complexity of included sentences, and as complexity increased, so did overall reading speeds. Perhaps surprisingly, on average, younger adults required an additional 50–100 ms to process syntactically complex sentences than older adults, and older adults' sentence reading was not affected by whether memory load nouns were matching or mismatching. Peculiarly, Kemper and Herman conclude this is evidence of older adults refusing to allocate sufficient resources for proper interpretation of complex structures. However, although Kemper and Herman gathered a large psychometric database, these measures were not included in any models for the results of the memory load task.

Kemper and Herman (2006), therefore, suggested that syntactic difficulty may interact with memory interference, and that interference effects may vary based on processing difficulty. The current study aimed to further investigate this possibility, focusing on memory and memory interference specifically. For older adults, manipulating memory interference is complicated by evidence suggesting that older adults place greater reliance on semantic information during reading. Evidence for this notion is presented by, for instance, Poulisse, Wheeldon, and Segaert (2019), who tested older and younger participants on subject‐verb agreement of minimally short sentences (e.g., “I cook”), in which memory demands were, therefore, kept to a minimum. Pseudoverbs were interspersed with real verbs, and while Poulisse et al. found that older adults were slower and less accurate than younger adults across conditions, older adults were further disproportionately less accurate on pseudowords compared to real words (see further Milburn et al., 2021). If older adults rely on semantic content more so than younger adults for successful interpretation of sentences, older groups could be more susceptible to similarity‐based interference than younger adults, contradicting results obtained by Kemper and Herman (2006). Nevertheless, to our knowledge, this has not been tested directly.

In short, significant gaps remain in the literature around older adults' processing of syntactic ambiguity and the ways in which changing memory dynamics affect such processing. Research on syntactic disambiguation and memory capacity paints a confusing picture: while Felser et al. (2003) found lower‐span participants to preferentially attach to NP2, evidence from Swets et al. (2007), Traxler (2009), and Payne et al. (2014) suggests the opposite. Kemper and Herman (2006) suggest memory interference effects may interact with syntactic preferences, but it is unclear how older adults' reduced memory capacity aligns with this notion. A combination of attachment disambiguation in aging with interference measures could, therefore, inform not only theories of aging, but also accounts of WM and language processing. This study is, to our knowledge, the first to attempt this combination. We aimed to address this question using an implicit self‐paced reading paradigm, and further manipulated memory interference using load words presented before each sentence (after Van Dyke & McElree, 2006). We predicted the following outcomes:

  • 1.

    Following results obtained by Swets et al. (2007) and Traxler (2009), we predict older adults may show a greater NP2 bias than younger adults or a less defined NP1 bias due to the accellerated decay of NP1 in older adults' memory systems (as reviewed in Salthouse, 1996) (in the present study, we define “bias” as reductions in reading speed for one type of bias compared to the other.)

  • 2.

    In the same vein, we predict greater effects of memory interference in older adults, as this group should struggle to maintain competing information in memory and retrieve it accurately and in a timely fashion. Additionally, the greater reliance older adults may place on semantic information, semantic interference should affect older adults more heavily (e.g., Gonzalez‐Burgos et al., 2019; Poulisse et al., 2019).

2. Methods

2.1. Participants

Participants took part in this study online through Prolific (Prolific, 2014). Younger participants (n = 35) ranged between 18 and 25 years of age, while older participants (n = 30) were all aged over 65. Five younger participants were eliminated from the data sample due to failing attentional checks or incomplete responses on one of the pre‐tests, resulting in a final sample size of 30. All participants confirmed they were native speakers of English before taking part, and were paid for their participation. All participants gave full informed consent before taking part, and the study received ethical approval from the University of Essex Social Sciences Ethics Sub‐committee.

Table 1 displays participant biodata in both groups. Education levels were measured along the International Standard Classification of Education (UNESCO Institute for Statistics, 2012). Groups did not differ on any of the measures below, with one exception: younger participants spent more years in education than older participants (t(44.5) = −4.26, p < .001), although differences in overall level of education were only marginally significant (t(47.7) = 1.91, p = .06).

Table 1.

Summary of participant biodata

Younger Group Older Group Comparison
Age M = 21.77; SD = 2.11 M = 68.53; SD = 3.44
Gender 21 Female, 9 Male 17 Female, 13 Male
Years in Education M = 16.57; SD = 1.66 M = 15.43; SD = 2.78 t(44.5) = −4.26, p < .001
Education Level Mode = 6 (Bachelor‐level) Mode = 5 (Short‐cycle Tertiary)
RST Score M = 23.33; SD = 5.96 M = 21.20; SD = 5.95 t(58) = 1.39, p = .171
LCT Score M = 28.03; SD = 5.90 M = 17.30; SD = 5.90 t(58) = 6.97, p < .001

Note. Education level was measured on the International Standard Classification of Education (UNESCO Institute for Statistics, 2012).

Abbreviations: M, mean; SD, standard deviation.

2.2. Materials

2.2.1. Pre‐tests

2.2.1.1. Reading Span Task

Participants completed the Reading Span Task (RST; Daneman & Carpenter, 1980) hosted using Qualtrics (Provo, UT, USA). Participants were shown sets of sentences incrementally increasing in number, ranging from three to eight sentences per set, and told to remember the final word of each sentence. Participants further rated sentences for appropriateness, with half of items designed as appropriate and half as inappropriate. The RST, therefore, accurately captures the concurrent storage and processing information involved in WM, and contrasts with tasks which assess only the storage aspect of cognition, such as simple digit span tasks. Ample work has addressed and confirmed the construct validity of the RST, and what RST scores actually reflect (see Daneman & Hannon, 2007, for a review). In the present study, participants saw each sentence individually before being asked to type the recall words into text boxes on a separate screen. Sentences used in the RST had an average length of 6.97 words (SD = 1.44, range = [5,11]). All sentences consisted of simple grammar and no technical jargon or phrases (see our Supplementary Materials at https://osf.io/eympz/ for a list of all RST items). Points were awarded to participants for recalling a correct word in the correct order, with half points given for correct words recalled in an incorrect text box in the same trial (following Conway et al., 2005). There were no significant by‐group differences on the RST (t(58) = −1.39, p = .171; M Y = 23.3, M O = 21.2). Age was not associated with different RST scores either when considering both groups together (r(58) = −.166, p = .206), or in the younger group (r(28) = .103, p = .591) or the older group considered separately (r(28) = .109, p = .566).

2.2.1.2. Letter Comparison Test

Processing Speed was assessed using the Letter Comparison Test (LCT; Salthouse, 1991; Salthouse & Babcock, 1991). Participants were presented with two letter strings of either three, six, or nine letters, and asked to judge whether these strings were identical or different. Scores were calculated as the amount of correctly identified strings within 30 s. Unlike commonly used Processing Speed tasks such as the digit symbol substitution task from the Wechsler Adult Intelligence Scale (WAIS; Wechsler, 2008), the LCT involves virtually no memory demands and is, therefore, a more direct method of assessing Processing Speed as confounding cognitive demands are kept to a minimum. Performance on the LCT is a reliable predictor of chronological age (Salthouse, 1991) and Processing Speed generally explains large portions of age‐related variance across studies (Madden, 2001). The LCT in this study comprised a maximum of 48 character pairs, equally subdivided into identical and different conditions. Strings were checked for wordlike items. The LCT was presented using PsychoPy 3 2020.1.3 and Pavlovia (Peirce & MacAskill, 2018; PsychoPy/Pavlovia, 2021): Strings appeared toward the left and right of the center of the screen, and participants completed five untimed practice trials before the main test. Group differences on the LCT were more pronounced than on the RST (see above; t(58) = 6.97, p < .001; M Y = 28.0, M O = 17.3); higher age was associated with lower Processing Speed across groups (r = −.65, p < .001). Further, RST and LCT scores were positively correlated in both the Younger (r = .16, p < .001) and Older groups (r = .12, p < .001), although this was not significant when participant groups were pooled (r = .157, p = .09).

2.2.2. Ambiguity condition

Materials for the main experiment were adapted from Traxler (2009), and included 120 fillers and 120 experimental sentences. All experimental items included a disambiguated relative clause with two possible antecedents, while filler items were constructed to match experimental items for length, and included complex grammar with some unambiguous relative clauses. Experimental items were equally subdivided into two bias conditions (60 trials per condition), NP1 (as in (1) below) and NP2 (2), which semantically biased participants to preferring one NP as the antecedent. Ambiguous relative clauses always began at the sixth word in the sentence, allowing for comparisons of reading times. The mean length of experimental items and fillers was 12.9 words, and NP1‐biased and NP2‐biased sentences were of equal length (t(118) = .00, p > .05; M NP1 = 12.3, M NP2 = 13.2).

  • (1)

    The owner of the house that had the moustache appealed for help.

  • (2)

    The owner of the house that needed renovation appealed for help.

To ensure our stimuli correctly biased participants to either ambiguity condition, we normed the 120 experimental sentences with a group of raters who did not take part in the main study, and who self‐declared they were native speakers of English (n = 13, M Age = 26.23, SDAge = 8.43, [49; 19]). We presented raters with the full list of experimental items and asked them to select the noun to which the relative clause in each sentence referred. On average, participants selected the intended referent in 98.6% of sentences (SD = 1.97, [94.77; 100]). Our stimuli, therefore, accurately captured NP biases. Full results of stimuli norming are presented in the Supplementary Materials at https://osf.io/eympz.

2.2.3. Similarity‐based interference

Interference was created by presenting three nouns before each trial sentence. A 2x2 design was created, combining interference with ambiguity condition: 30 trials in each ambiguity condition were matched with an interfering memory load, while the other 30 were preceded by a non‐interfering load. Participants were explicitly prompted to remember load words while reading (rather than referring to any words in the sentences they read), and faced recall prompts after the sentence in 50% of trials. Interference nouns were either semantically related or unrelated to actor(s) in the experimental sentences, since the actor NP must be retained in WM while processing is ongoing. For instance, in the examples above, interfering load items could include “mortgage,” “property,” and “residence,” while non‐interfering items could be “experiment,” “takeaway,” and “grill.” In filler sentences, interfering load NPs related to randomly selected elements in the sentence. Interfering and Non‐interfering sentences were of equal length (t(237.6) = −1.87, p > .05; M Interf. = 12.7, M Non‐interf. = 12.8). Load sets appear alongside corresponding sentences in the Supplementary Materials at https://osf.io/eympz/.

2.3. Procedure

Participants were informed they would be presented with three words they should try to remember, after which they would read a sentence and then recall one of the words they had attempted to remember. Each trial began with a 2000 ms fixation cross followed by the interference word set for 4000 ms. The duration of the load set was increased from Van Dyke and McElree's (2006) 3 s to allow for older participants, who may have required more time to fully memorize items (Jaroslawska & Rhodes, 2019; Salthouse & Babcock, 1991). All text was presented in the center of the screen in white Arial font on a gray background, using PsychoPy 2020.1.3 and Pavlovia (Peirce & MacAskill, 2018; PsychoPy/Pavlovia, 2021). Participants pressed the space bar to advance to word‐by‐word through sentences. Attention was measured by evaluating accuracy scores on comprehension questions and recall prompts. Accuracy was high across groups on questions (M = 89.5%) and prompts (M = 85.7%), and groups did not differ on comprehension accuracy (t(60.7) = −1.46, p > .05; M Y = 84.5, M O = 87.5) or recall (t(51.3) = 1.04, p > .05; M Y = 87.1, M O = 84.0). Fig. 1 displays the full sequence of a single trial item. Trials were randomly subdivided into four blocks of 60 sentences each, and participants were prompted to take a short break between blocks. The experiment took most participants between 55 and 70 min to complete.

Fig. 1.

Fig. 1

Example Trial sequence of a main experimental trial. Load word presentation time was extended to 4000 ms to allow for older readers' potentially slower storage. Recall prompts and comprehension questions were not mutually exclusive.

2.4. Analysis

Two Regions of Interest (ROIs) were defined: (1) a relative clause NP region of four words; and (2) a Spillover region of two words. For instance, in the sentence “The owner of the house that had the moustache appealed for help,” the NP ROI consists of “that had the moustache,” while the Spillover ROI comprises “appealed for.” Analysis was conducted in R (R Core Team, 2020) using lme4 for (generalized) linear mixed modeling (Bates et al., 2014), EMAtools for the generation of effect sizes (Kleiman, 2017), sjPlot for plotting models (Lüdecke, 2021), and emmeans for post‐hoc analysis (Lenth et al., 2019). Reading times were residualized by character count in each word to control, and Residualized Reading Times (RRTs) were then log‐transformed and outliers over 2.5 SDs above or below each participant's mean were trimmed, which resulted in a total of 12.9% of datapoints being removed prior to analysis. Pre‐test scores (RST and LCT) were centered before being added to any model.

We used maximal‐effects structures for mixed models for RRT and accuracy data: these included fixed effects for Ambiguity and Interference condition (and interactions of Ambiguity and Interference conditions, or either condition with pre‐tests), Age Group, centered RST Score, and centered LCT Score, as well as random intercepts for Participant and Trial whenever such intercepts could be included without compromising model convergence. Models for RRT data were fitted with a Gaussian (normal) distribution, while Accuracy models were fitted as generalized binomial mixed‐effects models. None of the (G)LMMs reported in this article included effects of education, as neither of the two collected measures of education improved model fit anywhere. Apart from analyzing RRTs, we also explored accuracy on the recall prompts included in the experimental sequence.

We further conducted Bayesian analyses to confirm the presence or absence of group‐related effects using the brms (Bürkner, 2017) library. We fitted exponential Gaussian distributions as weakly informative priors (as is typical of reaction times, see Matzke & Wagenmakers, 2009), included 3000 iterations per model, and contrasted models including an Age Group by Condition interaction term with models including only main effects of these terms (Wagenmakers, 2007). Model comparisons were conducted using the bridgesampling (Gronau et al., 2017) library, and resulted in Bayes' Factors (BFs). We adopted the conservative interpretation of Kass and Raftery (1995), and defined BFs of < 1 to signify strong support for the null model, while BFs of > 3 suggested evidence for the full model. Scripts used for analysis can be found in the Supplementary Materials.

3. Results

3.1. Ambiguity condition

Older readers read more slowly than the Younger group in both ROIs (ps < .01, ds < −0.8) and reading times did not vary by RST or LCT score (all ps > .05). The NP ROI did not capture ambiguity effects (t(6030) = −1.246, p > .05, d = −.032), and older and younger readers did not read either ambiguity faster than the other group (t(6029) = −.165, p > .05, d −.004). However, in the Spillover ROI, NP1‐biased items were read faster than NP2‐biased items (t(6057) = 2.396, p = .01, d = .062). Crucially, ambiguity preference did not vary by age group (t(6053) = −.457, p > .05). For full details of Ambiguity Condition models, please refer to Table 2. Bayesian model comparisons provided strong evidence for the null hypothesis; that is, for the absence of an Age Group by Ambiguity Condition interaction (Bayes FactorNP = 0961; BFSpillover = .0423; Lee & Wagenmakers, 2014).

Table 2.

Summary of Ambiguity Condition model in both ROIs

ROI Predictor Estimate SE T‐value p d
NP Intercept −2.400 .0070
Age Group −.3740 .1120 −3.340 .002 −.8849
LCT −.0095 .057 −.167 .8678 −.0443
RST .0017 .0428 .040 .9685 .0105
Ambiguity Condition −.0162 .0130 −1.246 .2126 −.0321
Ambiguity * Age Group −.0035 .0211 −.165 .8686 −.0043
Ambiguity * LCT .0016 .0108 .146 .8842 .0038
Ambiguity * RST −.0041 .0080 −.513 .6077 −.0132
Spillover Intercept −2.421 .0699
Age Group −.4032 .1115 −3.616 .0006 −.9509
LCT −.0127 .0568 −.224 .8238 −.0588
RST .0136 .4258 .319 .7512 .0839
Ambiguity Condition .0420 .0175 2.396 .0166 .0616
Ambiguity * Age Group −.0129 .0283 −.457 .6477 −.0117
Ambiguity * LCT .0138 .0145 .950 .3419 .0244
Ambiguity * RST −.0061 .0108 −.560 .5756 −.0144

Note. These models further included random intercepts for Trial (σ2 NP = .027, SDNP = .164; σ2 Spillover = .024, SDSpillover = .154) and Subject (σ2 NP = .100, SDNP = .317; σ2 Spillover = .098, SDSpillover = .314). Reference level for Ambiguity Condition was NP1‐biased. R2 NP – Conditional = .645, R2 NP – Marginal = .147; R2 Spillover – Conditional = .500, R2 Spillover – Marginal = .131.

Bolded values are those meeting the significance threshold of p < .05.

3.2. Interference condition

Interfering trials were read slower than Non‐interfering trials in the NP (t(6031) = 3.799, p < .001, d = .098) but not the Spillover ROI (t(6058) = −1.499, p > .05, d = −.039). Once again, older readers were slower across conditions in both ROIs (ps < .01, ds < −.8), but importantly, groups were equally affected by Interference manipulations in the NP (t(6029) = −.570, p > .05, d = .008) and Spillover ROI (t(6055) = .564, p > .05, d = .015) No pre‐test scores predicted reading times or interacted with Interference Condition in either ROI (ps > .05). Table 3 details model output for Interference models. As with Ambiguity models, comparison of Bayesian models resulted in BFNP = .0871 and BFSpillover = .1791, providing strong and moderate evidence for the null hypothesis, respectively.

Table 3.

Summary of Interference Condition model in both ROIs

ROI Predictor Estimate SE T‐value p d
NP Intercept −2.434 .0704
Age Group −.3790 .1120 −3.384 .0013 −.8958
LCT −.0055 .0571 −.097 .9235 −.0255
RST −.0072 .0429 −.170 .8658 −.0449
Interference Condition .0493 .0130 3.799 .0001 .0978
Interference * Age Group .0065 .0209 .309 .7570 .0080
Interference * LCT −.0061 .0107 −.570 .5685 −.0147
Interference * RST .0135 .0080 1.684 .0923 .0434
Spillover Intercept −2.387 .0700
Age Group −.4184 .0117 −3.747 .0004 −.9841
LCT .0050 .0569 .088 .9305 .0230
RST .0070 .0426 .180 .8574 .0474
Interference Condition −.0263 .0175 −1.499 .1339 −.0385
Interference * Age Group .0160 .0283 .564 .5725 .0145
Interference * LCT −.0211 .0144 −1.461 .1442 −.0376
Interference * RST .0062 .0108 .575 .5651 .0148

Note. These models further included significant random intercepts for Trial (σ2 NP = .027, SDNP = .164; σ2 Spillover = .237, SDSpillover = .154) and Subject (σ2 NP = .100, SDNP = .317; σ2 Spillover = .098, SDSpillover = .314). Reference level for Interference Condition was Non‐Interfering. R2 NP – Conditional = .647, R2 NP – Marginal = .150; R2 Spillover – Conditional = .500, R2 Spillover – Marginal = .131.

Bolded values are those meeting the significance threshold of p < .05.

We also fitted models including an interaction between Ambiguity and Interference condition. In the NP ROI, this interaction showed reversed ambiguity preferences depending on Interference condition t(6029) = −9.229, p < .001, d = −.238): under non‐interfering memory loads, NP1 was preferred, while NP2‐biased trials were read faster under Interfering loads. The two age groups did not respond differently to this interaction (t(6023) = −.150, p > .05, d = −.004). The interaction between Ambiguity and Interference Condition was not replicated in the Spillover ROI (t(6055) = −1.508, p > .05, d = −.012). Interaction models are further detailed in Table 4.

Table 4.

Summary of Condition Interaction model in both ROIs

ROI Predictor Estimate SE T‐value p d
NP Intercept −2.478 .0707
Age Group −.3753 .1122 −3.344 .0015 −.8810
LCT −.0091 .0568 −.160 .8735 −.0427
RST −.0008 .0043 −.0197 .9846 −.0052
Interference Condition .1493 .0152 9.849 <.001 .2536
Ambiguity Condition .0865 .0155 5.595 <.001 .1441
Interference * Ambiguity −.1988 .0215 −9.229 <.001 −.2377
Interference * Age Group .0052 .0214 .246 .8055 .0062
Ambiguity * Age Group −.0017 .0218 −.000 .9366 −.0020
Interference * Ambiguity * Age Group −.0046 .0030 −.150 .8811 −.0039
Spillover Intercept −2.414 .0707
Age Group −.3889 .1121 −3.469 .0010 −.9020
LCT −.0056 .0563 −.099 .9214 −.2649
RST .0106 .0423 .252 .8023 .0672
Interference Condition −.0058 .0209 −0.279 .7806 −.0072
Ambiguity Condition .0404 .0213 1.897 .0579 .0489
Interference * Ambiguity −.0133 .0296 −.452 .6516 −.0116
Interference * Age Group −.0443 .0294 −1.508 .1317 −.0388
Ambiguity * Age Group −.3130 .0299 −1.046 .2958 −.0269
Interference * Ambiguity * Age Group .0681 .0417 1.635 .1021 .0420

Note. These models further included significant random intercepts for Trial (σ2 NP = .027, SDNP = .163; σ Spillover = .024, SDSpillover = .154) and Subject (σ NP = .101, SDNP = .32; σ Spillover = .098, SDSpillover = .314). Reference levels were NP1‐biased for Ambiguity Condition and Non‐Interfering for Interference Condition. R2 NP – Conditional = .657, R2 NP – Marginal = .160; R2 Spillover – Conditional = .501, R2 Spillover – Marginal = .132.

Bolded values are those meeting the significance threshold of p < .05.

3.3. Recall prompts

We further examined accuracy and reaction times (RTs) on recall prompts, which were presented on 50% of trials. While recall accuracy was not affected by Ambiguity Condition as a main effect (t(103) = .233, p > .05, d = .006), readers with higher WM spans showed marginally higher recall accuracy on dispreferred NP2 structures (t(7146) = 1.926, p = .05, d = .045). This interaction is further visualized in Fig. 3. No effects of pre‐test scores were found (all ps > .05). Similarly, recall correctness was unaffected by Interference Condition (t(7136) = .021, p > .05, d ≈ .000). But, in Interference models, high‐span participants showed higher accuracy on recall prompts (t(10) = 2.770, p = .006, d = .533). Accuracy scores did not differ by age or LCT score (ps > .05). Recall accuracy models are fully detailed in Table 5.

Fig. 3.

Fig. 3

Plot showing the significant impact of RST score on recall prompt accuracy, the explicit measure recorded in this study. Higher‐span readers were additionally facilitated on NP2‐biased trials, which were dispreferred in both age groups.

Table 5.

Summary of generalized binomial models predicting Recall Prompt Accuracy

Condition Predictor Estimate SE Z‐value p d
Ambiguity Intercept .−.362 .078
Age Group .151 .125 1.210 .226 .231
LCT −.075 .063 −1.178 .239 −.228
RST .066 .048 1.396 .163 .267
Ambiguity Condition .019 .081 .233 .816 .006
Ambiguity * Age Group −.106 .125 −.808 .419 −.020
Ambiguity * LCT .024 .066 .356 .722 .009
Ambiguity * RST .096 .050 1.926 .054 .045
Interference Intercept −.352 .078
Age Group .790 .12 .633 .527 .116
LCT −.080 .064 −1.261 .207 −.243
RST .132 .048 2.770 .006 .533
Interference Condition .002 .081 .021 .983 .000
Interference * Age Group .038 .131 .293 .769 .007
Interference * LCT .034 .066 .518 .604 .012
Interference * RST −.036 .050 −.718 .473 −.017

Note. These models further included significant random effects for Trial (σ2 Ambiguity = .004, SDAmbiguity = .066; σ2 Interference = .003, SDInterference = .059) and Subject (σ2 Ambiguity = .057, SDAmbiguity = .240; σ2 Interference = .057, SDInterference = .240). Reference level for Ambiguity Condition was NP1‐biased, reference level for Interference Condition was Interfering. R2 Conditional – Ambig. = .019; R2 Marginal – Interf. = .004; R2 Conditional – Interf. = .018; R2 Marginal – Interf. = .003.

Bolded values are those meeting the significance threshold of p < .05.

4. Discussion

The present study combined measurements of relative clause attachment preferences with similarity‐based memory interference conditions in older and younger adults. Younger (18–25) and older (65+) adults read sentences with NP1‐ and NP2‐biased relative clauses and while remembering interfering or non‐interfering three‐word memory loads. This study was the first to combine memory load interference with disambiguation in groups of older and younger adults.

4.1. Disambiguation bias and memory interference

Our research shows robust evidence for an NP1 bias in both older and younger readers, contradicting the Recency strategy (Li & Sheng, 2017; Sturt et al., 2002). Recency preferences have been widely reported in both older (e.g., Cuetos & Mitchell, 1988) and more recent (e.g., Akal, 2021) investigations, yet this is not the only reading study to report an NP1‐preference: both Swets et al. (2007) and Traxler (2009) reported participants generally preferred attachment to NP1. Although these studies intentionally segmented sentences to elicit NP1‐biases, which we did not, participants in our study were allowed to set their own reading speed, which would have allowed for segmentation at any point. Perhaps more crucially, previous research suggests that variations in animacy of potential referents can bias speakers and readers to preferentially attach clauses to the animate referent (for discussions, see Kwon et al., 2019; Mak et al., 2006; Traxler et al., 2005). We did not actively control for animacy in our stimuli, and as such, the overall preference for NP1 we discovered may not be entirely surprising.

Crucially, we discovered that under non‐interfering memory loads, both groups preferred NP1 attachment, while interfering loads resulted in a preference for NP2. The lack of recent implicit behavioral results in similar experiments make an informed interpretation of these results difficult. However, our findings align at least to some extent with those of Kemper and Herman (2006), who found greater memory interference (in younger adults) on the most complex of grammatical structures (see also Gordon et al., 2002, for similar findings). In our study, NP2‐biased items were the structure which required greater time and possibly greater resources, at least under non‐interfering conditions. Our results, therefore, mirror those of Kemper and Herman (2006), in concluding that interference was exacerbated on syntactic structures which require greater processing capacity, or more processing resources. Alternatively, rather than greater interference effects in NP2‐biased trials, it could be the case that interference was in fact greater in NP1‐biased sentences, and that NP2 bias acts as a “default” interpretation for when processing load increases. This could be evidenced by the patterns observed in Fig. 2: reading times on NP2‐biased items are seemingly less affected by interference manipulations, while greater differences are observed on N1‐biased sentences in both age groups. Future research should elucidate what structures and interpretations are specifically more susceptible to interference manipulations.

Fig. 2.

Fig. 2

Model plot of residualized reading times in the NP ROI by Ambiguity and Interference Condition, displaying the reversal of Ambiguity preferences depending on interfering or non‐interfering memory load.

Generally, we suggest this interaction of ambiguity and interference reflects a stacking of constraints on the linguistic processor, adding to constraint‐based models of sentence processing which take and integrate information from many different sources (e.g., Futrell et al., 2020; McRae & Matsuki, 2013; Van Gompel, 2013). Reading non‐preferred NP2‐biased structures (the first constraint in this study) significantly increased reading times. The introduction of interference did not simply negate or confound the effects of the initial constraint (disambiguation), but rather added to the processing difficulty the first constraint presented. In our study, reading times R may, therefore, be predicted by an additive function of disambiguation A and interference I: RRTs∼AAmbiguity+IInterference. Adding additional linguistic constraints (such as grammatical complexity, predictiveness, &c.) could add to this formula: RRTs∼AAmbiguity+IInterference+CComplexity+PProbability+⋯+⋯ In future studies, it should be established whether other constraints stack in the same way, and whether constraint stacking results in a linear, exponential, or approximant‐style curve.

Alternatively, stacked constraints may be connected to Salthouse's (1996) processing speed theory of cognition. In the Salthouse account, information processed early on in the processing sequence of low‐speed readers is lost by the time it is necessary for later‐stage processing, leading to decreased quality of processing, which then leads to less effective retrieval at later stages, and so forth. Adding stacked constraints to this vicious cycle could have the potential to additionally delay early stages of processing and/or the quality of processed information, leading to greatly reduced processing effectiveness. Given our current data, we consider this interpretation unlikely given the lack of statistical impact of LCT scores on our effects of interest, as well as older adults' similar ambiguity and interference effects despite significantly lower performance on the LCT compared to younger adults. We nevertheless consider this a highly worthwhile avenue for future research.

Notably, the interaction between interference and ambiguity condition occurred in both age groups. Older and Younger adults were, therefore, not only equally susceptible to interference in general, but the effects of memory interference on disambiguation patterns were also age‐invariant. The quality of WM operations, therefore, did not selectively affect older readers' linguistic processing. We further consider it unlikely that compensation strategies in older adults resulted in this lack of age differences: one of the most well‐documented compensatory mechanisms, in which older adults may rely more heavily on semantic compared to syntactic information (e.g., Poulisse et al., 2019), is hard to reconcile with this study's similarity‐based memory interference findings.

Poulisse et al. (2019) found older adults placed greater reliance on semantic information during syntax processing, and off‐line performance declined more severely in older adults when semantic context was absent than in younger adults. Further, results from semantic priming studies suggest that adults' sensitivity to semantic information becomes more refined with age—indeed, Laver and Burke (1993) suggest semantic priming effects increase as age progresses, in line with the increased reliance on semantic information proposed by Poulisse et al. (2019). Nevertheless, if older adults rely more on semantic information during processing, and struggle to process semantically impoverished stimuli (as Poulisse et al. claim), memory interference based on semantic information should impact older adults more severely. This was not the case in the present study. We, therefore, suggest the quality of memory operations did not decline with age in our sample.

This study also contradicts the results of Payne et al. (2014), one of the few examinations of disambiguation preferences in older adults, as we did not find that disambiguation preferences were modulated by WM or age. Our results could differ from the findings of Payne et al. due to the absence of by‐group differences on the RST in this study: both groups showed roughly equal WM scores (M Y = 23.3, M O = 21.2). Nevertheless, greater age was associated with lower RST scores across groups and within both groups in our study, and the effects of RST score in recall prompt models suggest our RST design was effective.

It could also be the case that our measure of Processing Speed, the LCT, accounted for much of the variance Payne et al. accredit to WM span. LCT Score correlated heavily with age in our study and group differences were far more pronounced than on the RST. Furthermore, past studies have emphasized the interplay between Processing Speed and WM: for instance, Salthouse (1992) found that controlling for Processing Speed took away much memory‐related variance in an age‐continuous sample of participants. Additionally, both Processing Speed and WM are subserved by similar cognitive resources or even neural structures, both in development (Fry & Hale, 2000; Newbury, Klee, Stokes, & Moran, 2016) and decline (Kim & Park, 2018; Unibaso‐Markaida, Iraurgi, Ortiz‐Marqués, Amayra, & Martínez‐Rodríguez, 2019).

We further emphasize the dissociation between implicit and explicit memory, especially in older adults (see further Waters & Caplan, 2001; Bopp & Verhaeghen, 2005; Hicks, Alexander, & Bahr, 2018; though cf. Ward & Shanks, 2018). Our findings from recall prompts, which do require conscious memory recall, suggest that distinguishing explicit and implicit memory is key to uncovering language processing differences between age groups. Notably, recall accuracy marginally interacted with our measure of WM, which showed that high‐span participants across groups exhibited less difficulty answering prompts on NP1‐biased items than their lower‐span peers. However, this interaction was not found in the analysis of reading times. This dissociation may at least partially be due to the varying cognitive and linguistic underpinnings of our measurements (ambiguity resolution compared to memory recall), and future studies should contrast implicit and explicit measures of the same construct to fully validate this notion. Nevertheless, the absence of WM effects on implicit measures of this study contradicts previous research supporting the notion that individual differences in WM capacity affect syntax processing Just and Carpenter (1992); Felser et al. (2003); Yoo et al. (2017), and instead suggest these effects are dependent on explicit task and memory demands.

This study is not alone in suggesting minimal impact of WM on disambiguation measures, however. Both Evans et al. (2015) and Traxler (2009) suggest the impact of WM on processes underlying disambiguation is minimal. The conflict between our data and Payne et al.'s (2014) is, therefore, not unique. We suggest that methodological differences are at least partially responsible for this dissociation: similar to age‐only effects, WM limitations may affect explicit tasks more than implicit measures—as our additional analysis of recall prompts showed—and this distinction often causes contradictory conclusions about older adults' memory and language functions. Indeed, this notion is supported by recent studies on syntax processing in aging by Hardy, Messenger, & Maylor (2017, 2020) and Van Boxtel and Lawyer (2022), which found intact implicit processing patterns in older adults unaffected by WM scores. More broadly, our study, therefore, makes the case for a significant nuancing of the view that syntax processing declines with age.

This study included several limitations which we acknowledge and emphasize for future research. First, we used different sentence exemplars between the NP1‐ and NP2‐biased conditions. We deemed it necessary to design the study in this way to most effectively examine individual differences between participants and evaluate the interaction between Ambiguity and Interference. However, this methodological choice does raise the possibility that differences between exemplars confounded bias conditions. This is not a possibility we completely discount, and follow‐up studies may want to consider using the same exemplars in both bias conditions (e.g., after Traxler, 2009). Second, we chose to compute NP and Spillover regions to evaluate reading times, which could have limited the scope of our analyses and could have interfered with end‐of‐sentence wrap‐up effects and effects resulting from the length of each relative clause. Our NP and Spillover regions were not defined according to the length of each clause, but rather set generally across sentence types. Intervening constituents or words may, therefore, have inadvertently affected our results. Although we partially controlled for length effects by RRTs, we acknowledge we did not fully eliminate the possibility for interfering sentence length effects. Future studies may want to control their stimuli more carefully in this regard.

Finally, our participants were sampled from Prolific, an online database, which may be more prone to “dishonest” or “fake” participants (e.g., Roehl & Harland, 2022). It is a possibility that some participants were not truthful in their self‐reports. While we cannot fully discount this possibility, there are several reasons to believe our online samples were reliable and not dishonest. First, our strong age differences on the LCT suggest participants honestly reported their chronological age. Second, the online recruitment system we used, Prolific.co (Prolific, 2014), is consistently rated the most reliable online participant platform with among the highest rates of attention, comprehension, honesty, and reliability (Eyal, David, Andrew, Zak, & Ekaterina, 2022; Gagné & Franzen, 2023). Prolific not only includes an identity verification system where prospective participants upload photo ID before taking a photo of their face (with Prolific matching the participant to their ID) but it also tracks completed, attempted, and abandoned research attempts, so that researchers can exclude data from quickly aborted or repeated sessions. Indeed, for the present study, the total number of participants who attempted the study online was upward of 100, even though our final sample size was 65. Although it is impossible to completely rule out the possibility of our sample including dishonest participants, we are optimistic that included participants reported their correct chronological age and were attentive to the study.

5. Conclusion

In short, this study was the first to combine similarity‐based interference with syntactic disambiguation in older adults. Using a self‐paced reading paradigm, we discovered age‐invariant disambiguation and memory interference effects. All participants were affected by an interaction of interference and ambiguity, such that stronger interference‐related slow‐downs of reading speed were found in dispreferred NP2‐biased sentences in both groups. Our results support processing models which emphasize the mutual interaction of memory and language processing, and make the case for further, large‐scale research into this topic with various methodologies.

Conflict of interest statement

The authors have no conflict of interest to disclose.

Supporting information

Data S1 

TOPS-18-0-s001.pdf (100.7KB, pdf)

Acknowledgments

The authors would like to thank the participants who took part in this study.

This article is part of the topic “Aging Language: Understanding How Human Cognition Changes As We Grow Old,” Antonio Benítez‐Burraco and Olga Ivanova (Topic Editors).

References

  1. Abney, S. P. (1991). Parsing by chunks. In Robert C., Berwick , Abney Steven P., & Tenny Carol (Eds.), Principle‐based parsing: Computation and Psycholinguistics (pp. 257–278). Springer. [Google Scholar]
  2. Akal, T. (2021). Recency preference in ambiguous relative clause attachment in Turkish. Journal of Language and Linguistic Studies, 17, 139–159. [Google Scholar]
  3. Altmann, G. T. , van Nice, K. Y. , Garnham, A. , & Henstra, J.‐A. (1998). Late closure in Context. Journal of Memory and Language, 38(4), 459–484. [Google Scholar]
  4. Bates, D. , Mächler, M. , Bolker, B. , & Walker, S. (2014). Fitting linear mixed‐effects models using lme4. arXiv preprint arXiv:1406.5823.
  5. Beese, C. , Meyer, L. , Werkle‐Bergner, M. , Düzel, S. , Kühn, S. , Demith, I. , Lindenberger, U. , & Friederici, A. D. (2017). The benefit of syntactic chunking for language comprehension in the aging brain. In 7th IMPRS Neurocom Summer School.
  6. Bopp, K. L. , & Verhaeghen, P. (2005). Aging and verbal memory span: A meta‐analysis. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 60(5), P223–P233. [DOI] [PubMed] [Google Scholar]
  7. Bürkner, P.‐C. (2017). brms: An R package for Bayesian multilevel models using Stan. Journal of Statistical Software, 80(1), 1–28. [Google Scholar]
  8. Carreiras, M. , & Clifton Jr, C. (1993). Relative clause interpretation preferences in Spanish and English. Language and Speech, 36(4), 353–372. [DOI] [PubMed] [Google Scholar]
  9. Checa‐Garcia, I. (2016). Prosodic cues in relative clauses disambiguation: Bilinguals vs. L2 learners. Journal of New Approaches in Educational Research, 5(2), 74–80. [Google Scholar]
  10. Conway, A. R. , Kane, M. J. , Bunting, M. F. , Hambrick, D. Z. , Wilhelm, O. , & Engle, R. W. (2005). Working memory span tasks: A methodological review and user's guide. Psychonomic Bulletin & Review, 12(5), 769–786. [DOI] [PubMed] [Google Scholar]
  11. Cuetos, F. , & Mitchell, D. C. (1988). Cross‐linguistic differences in parsing: Restrictions on the use of the late closure strategy in Spanish. Cognition, 30(1), 73–105. [DOI] [PubMed] [Google Scholar]
  12. Daneman, M. , & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Memory and Language, 19(4), 450. [Google Scholar]
  13. Daneman, M. , & Hannon, B. (2007). What do working memory span tasks like reading span really measure. In Osaka Naoyuki, Logie Robert H., & D'Esposito Mark (Eds.), The Cognitive Neuroscience of Working Memory (pp. 21–42). Oxford University Press. [Google Scholar]
  14. Evans, W. S. , Caplan, D. , Ostrowski, A. , Michaud, J. , Guarino, A. J. , & Waters, G. (2015). Working memory and the revision of syntactic and discourse ambiguities. Canadian Journal of Experimental Psychology/Revue Canadienne de Psychologie Expérimentale, 69(1), 136. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Eyal, P. , David, R. , Andrew, G. , Zak, E. , & Ekaterina, D. (2022). Data quality of platforms and panels for online behavioral research. Behavior Research Methods, 54, 1–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Felser, C. , Marinis, T. , & Clahsen, H. (2003). Children's processing of ambiguous sentences: A study of relative clause attachment. Language Acquisition, 11(3), 127–163. [Google Scholar]
  17. Frazier, L. , & Fodor, J. D. (1978). The sausage machine: A new two‐stage parsing model. Cognition, 6(4), 291–325. [Google Scholar]
  18. Frazier, L. , & Rayner, K. (1982). Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14(2), 178–210. [Google Scholar]
  19. Fry, A. F. , & Hale, S. (2000). Relationships among processing speed, Working Memory, and fluid intelligence in children. Biological Psychology, 54(1–3), 1–34. [DOI] [PubMed] [Google Scholar]
  20. Futrell, R. , Gibson, E. , & Levy, R. P. (2020). Lossy‐context surprisal: An information‐theoretic model of memory effects in sentence processing. Cognitive Science, 44(3), e12814. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gagné, N. , & Franzen, L. (2023). How to run behavioural experiments online: Best practice suggestions for cognitive psychology and neuroscience. Swiss Psychology Open, 3(1), 1‐21.
  22. Gibson, E. , Pearlmutter, N. , Canseco‐González, E. , & Hickok, G. (1996). Cross‐linguistic attachment preferences: Evidence from English and Spanish. Cognition, 59, 23–59. [DOI] [PubMed] [Google Scholar]
  23. Gonzalez‐Burgos, L. , Hernández‐Cabrera, J. A. , Westman, E. , Barroso, J. , & Ferreira, D. (2019). Cognitive compensatory mechanisms in normal aging: A study on verbal fluency and the contribution of other cognitive functions. Aging (Albany NY), 11(12), 4090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Gordon, P. C. , Hendrick, R. , & Johnson, M. (2001). Memory interference during language processing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(6), 1411. [DOI] [PubMed] [Google Scholar]
  25. Gordon, P. C. , Hendrick, R. , Johnson, M. , & Lee, Y. (2006). Similarity‐based Interference during language comprehension: Evidence from eye tracking during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32(6), 1304. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Gordon, P. C. , Hendrick, R. , & Levine, W. H. (2002). Memory‐load interference in syntactic processing. Psychological Science, 13(5), 425–430. [DOI] [PubMed] [Google Scholar]
  27. Gronau, Q. F. , Singmann, H. , & Wagenmakers, E.‐J. (2017). bridgesampling: An R package for estimating normalizing constants. arXiv preprint arXiv:1710.08162.
  28. Hardy, S. M. , Messenger, K. , & Maylor, E. A. (2017). Aging and syntactic representations: Evidence of preserved syntactic priming and lexical boost. Psychology and Aging, 32(6), 588. [DOI] [PubMed] [Google Scholar]
  29. Hardy, S. M. , Wheeldon, L. , & Segaert, K. (2020). Structural priming is determined by global syntax rather than internal phrasal structure: Evidence from young and older adults. Journal of Experimental Psychology: Learning, Memory, and Cognition, 64(4), 720–740. [DOI] [PubMed] [Google Scholar]
  30. Hicks, R. E. , Alexander, V. E. , & Bahr, M. (2018). Explicit and implicit memory loss in aging. International Journal of Pharmaceutical Sciences, 10(3), 40–52. [Google Scholar]
  31. James, A. N. , Fraundorf, S. H. , Lee, E.‐K. , & Watson, D. G. (2018). Individual differences in syntactic processing: Is there evidence for reader–text interactions? Journal of Memory and Language, 102, 155–181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Jaroslawska, A. J. , & Rhodes, S. (2019). Adult age differences in the effects of processing on storage in working memory: A meta‐analysis. Psychology and Aging, 34(4), 512. [DOI] [PubMed] [Google Scholar]
  33. Just, M. A. , & Carpenter, P. A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99(1), 122. [DOI] [PubMed] [Google Scholar]
  34. Kass, R. E. , & Raftery, A. E. (1995). Bayes factors. Journal of the American Statistical Association, 90(430), 773–795. [Google Scholar]
  35. Kemper, S. , & Herman, R. E. (2006). Age differences in memory‐load interference effects in syntactic processing. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 61(6), P327–P332. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Kemtes, K. A. , & Kemper, S. (1997). Younger and older adults' on‐line processing of ayntactically ambiguous sentences. Psychology and Aging, 12(2), 362. [DOI] [PubMed] [Google Scholar]
  37. Kim, S.‐J. , & Park, E. H. (2018). Relationship of Working Memory, processing speed, and fluid reasoning in psychiatric patients. Psychiatry Investigation, 15(12), 1154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Kleiman, E. (2017). Ematools: Data management tools for real‐time monitoring/ecological momentary assessment data. R Package Version 0.1.3.
  39. Kwon, N. , Ong, D. , Chen, H. , & Zhang, A. (2019). The role of animacy and structural information in relative clause attachment: Evidence from Chinese. Frontiers in Psychology, 10, 1576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Laver, G. D. , & Burke, D. M. (1993). Why do semantic priming effects increase in old age? A meta‐analysis. Psychology and Aging, 8(1), 34. [DOI] [PubMed] [Google Scholar]
  41. Lüdecke, D. (2021). sjplot ‐ Data visualization for statistics in social science. R Package.
  42. Lee, M. D. , & Wagenmakers, E.‐J. (2014). Bayesian cognitive modeling: A practical course. Cambridge University Press. [Google Scholar]
  43. Lemaire, B. , Pageot, A. , Plancher, G. , & Portrat, S. (2018). What is the time course of working memory attentional refreshing? Psychonomic Bulletin & Review, 25(1), 370–385. [DOI] [PubMed] [Google Scholar]
  44. Lemaire, B. , & Portrat, S. (2018). A computational model of working memory integrating time‐based decay and interference. Frontiers in Psychology, 9, 416. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Lenth, R. , Buerkner, P. , Herve, M. , Love, J. , Riebl, H. , & Singmann, H. (2019). Estimated marginal means, aka least‐squares means. R Package Version 1.3. 5.1.
  46. Lewis, R. L. , Vasishth, S. , & Van Dyke, J. A. (2006). Computational principles of working memory in sentence comprehension. Trends in Cognitive Sciences, 10(10), 447–454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Li, H. , & Sheng, X. (2017). A study on the garden path phenomenon from the perspective of generative grammar. Journal of Language Teaching and Research, 8(6), 1190–1194. [Google Scholar]
  48. Madden, D. J. (2001). Speed and timing of behavioral processes. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (pp. 288‐312). San Diego, CA: Academic Press.
  49. Mak, W. M. , Vonk, W. , & Schriefers, H. (2002). The influence of animacy on relative clause processing. Journal of Memory and Language, 47(1), 50–68. [Google Scholar]
  50. Mak, W. M. , Vonk, W. , & Schriefers, H. (2006). Animacy in processing relative clauses: The hikers that rocks crush. Journal of Memory and Language, 54(4), 466–490. [Google Scholar]
  51. Martin, N. , Kohen, F. , & Kalinyak‐Fliszar, M. (2010). A processing approach to the assessment of language and verbal short‐term memory in aphasia.
  52. Matzke, D. , & Wagenmakers, E.‐J. (2009). Psychological interpretation of the ex‐Gaussian and shifted Wald parameters: A diffusion model analysis. Psychonomic Bulletin & Review, 16(5), 798–817. [DOI] [PubMed] [Google Scholar]
  53. Mayer, J. F. , & Murray, L. L. (2012). Measuring working memory deficits in aphasia. Journal of Communication Disorders, 45(5), 325–339. [DOI] [PubMed] [Google Scholar]
  54. McRae, K. , & Matsuki, K. (2013). Constraint‐based models of sentence processing. Sentence Processing, 519, 51–77. [Google Scholar]
  55. Milburn, E. , Dickey, M. W. , Warren, T. , & Hayes, R. (2021). Increased reliance on world knowledge during language comprehension in healthy aging: Evidence from verb‐argument prediction. Aging, Neuropsychology, and Cognition, 30(1), 1–34. [DOI] [PMC free article] [PubMed]
  56. Newbury, J. , Klee, T. , Stokes, S. F. , & Moran, C. (2016). Interrelationships between Working Memory, processing speed, and language development in the age range 2–4 years. Journal of Speech, Language, and Hearing Research, 59(5), 1146–1158. [DOI] [PubMed] [Google Scholar]
  57. Oberauer, K. , & Lin, H.‐Y. (2017). An interference model of visual working memory. Psychological Review, 124(1), 21. [DOI] [PubMed] [Google Scholar]
  58. Payne, B. R. , Grison, S. , Gao, X. , Christianson, K. , Morrow, D. G. , & Stine‐Morrow, E. A. (2014). Aging and individual differences in binding during sentence understanding: Evidence from temporary and global syntactic attachment ambiguities. Cognition, 130(2), 157–173. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Peirce, J. , & MacAskill, M. (2018). Building experiments in PsychoPy. Sage. [Google Scholar]
  60. Poulisse, C. , Wheeldon, L. , & Segaert, K. (2019). Evidence against preserved syntactic comprehension in healthy aging. Journal of Experimental Psychology: Learning, Memory, and Cognition, 45(12), 2290–2308. [DOI] [PubMed] [Google Scholar]
  61. Prolific . (2014). Prolific. Available at https://www.prolific.co/.
  62. PsychoPy/Pavlovia . (2021). Pavlovia: Where behaviour is studied. Open SCience Tools Ltd. Available from https://pavlovia.org/.
  63. Qualtrics . (2021). Qualtrics Survey Software. Provo, UT: Qualtrics. Available from https://www.qualtrics.com.
  64. R Core Team . (2020). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. [Google Scholar]
  65. Roehl, J. M. , & Harland, D. J. (2022). Imposter participants: Overcoming methodological challenges related to balancing participant privacy with data quality when using online recruitment and data collection. Qualitative Report, 27(11), 2469–2485. [Google Scholar]
  66. Salthouse, T. A. (1991). Mediation of adult age differences in cognition by reductions in working memory and speed of processing. Psychological Science, 2(3), 179–183. [Google Scholar]
  67. Salthouse, T. A. (1992). Influence of processing speed on adult age differences in Working Memory. Acta Psychologica, 79(2), 155–170. [DOI] [PubMed] [Google Scholar]
  68. Salthouse, T. A. (1996). The processing‐speed theory of adult age differences in cognition. Psychological Review, 103(3), 403. [DOI] [PubMed] [Google Scholar]
  69. Salthouse, T. A. (2000). Aging and measures of processing speed. Biological Psychology, 54(1–3), 35–54. [DOI] [PubMed] [Google Scholar]
  70. Salthouse, T. A. , & Babcock, R. L. (1991). Decomposing adult age differences in working memory. Developmental Psychology, 27(5), 763. [Google Scholar]
  71. Salthouse, T. A. , & Coon, V. E. (1993). Influence of task‐specific processing speed on age differences in memory. Journal of Gerontology, 48(5), P245–P255. [DOI] [PubMed] [Google Scholar]
  72. Sturt, P. , Scheepers, C. , & Pickering, M. (2002). Syntactic ambiguity resolution after initial misanalysis: The role of recency. Journal of Memory and Language, 46(2), 371–390. [Google Scholar]
  73. Swets, B. , Desmet, T. , Hambrick, D. Z. , & Ferreira, F. (2007). The role of working memory in syntactic ambiguity resolution: A psychometric approach. Journal of Experimental Psychology: General, 136(1), 64. [DOI] [PubMed] [Google Scholar]
  74. Traxler, M. J. (2009). A hierarchical linear modeling analysis of working memory and implicit prosody in the resolution of adjunct attachment ambiguity. Journal of Psycholinguistic Research, 38(5), 491–509. [DOI] [PubMed] [Google Scholar]
  75. Traxler, M. J. , Williams, R. S. , Blozis, S. A. , & Morris, R. K. (2005). Working memory, animacy, and verb class in the processing of relative clauses. Journal of Memory and Language, 53(2), 204–224. [Google Scholar]
  76. UNESCO Institute for Statistics . (2012). International Standard Classification of Education: ISCED 2011. Montreal: UNESCO Institute for Statistics. [Google Scholar]
  77. Unibaso‐Markaida, I. , Iraurgi, I. , Ortiz‐Marqués, N. , Amayra, I. , & Martínez‐Rodríguez, S. (2019). Effect of the Wii Sports Resort on the improvement in attention, processing speed and Working Memory in moderate stroke. Journal of Neuroengineering and Rehabilitation, 16(1), 1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  78. Van Boxtel, W. S. , & Lawyer, L. A. (2022). Syntactic comprehension priming and lexical boost effects in older adults. Language, Cognition and Neuroscience, 38(1), 1–16. [Google Scholar]
  79. Van Dyke, J. A. , & McElree, B. (2006). Retrieval interference in sentence comprehension. Journal of Memory and Language, 55(2), 157–166. [DOI] [PMC free article] [PubMed] [Google Scholar]
  80. Van Gompel, R. P. (2013). Sentence processing. Psychology Press. [Google Scholar]
  81. Wagenmakers, E.‐J. (2007). A practical solution to the pervasive problems of p values. Psychonomic Bulletin & Review, 14(5), 779–804. [DOI] [PubMed] [Google Scholar]
  82. Ward, E. V. , & Shanks, D. R. (2018). Implicit memory and cognitive aging. In Oxford Research Encyclopedia of Psychology.
  83. Waters, G. S. , & Caplan, D. (1996). The capacity theory of sentence comprehension: Critique of Just and Carpenter (1992). [DOI] [PubMed]
  84. Waters, G. S. , & Caplan, D. (2001). Age, working memory, and on‐line syntactic processing in sentence comprehension. Psychology and Aging, 16(1), 128. [DOI] [PubMed] [Google Scholar]
  85. Wechsler, D. (2008). Wechsler Adult Intelligence Scale – Fourth edition. Pearson.
  86. Werner, L. L. (2019). Forgetting distractors: Inhibition or decay? PhD thesis, University of Nevada, Las Vegas. [Google Scholar]
  87. Yang, Y. , Chen, M. , He, W. , & Merrill, E. C. (2020). The role of working memory in implicit memory: A developmental perspective. Cognitive Development, 55, 100–129.
  88. Yoo, H. , Dickey, M. W. , Yoo, H. , & Dickey, M. W. (2017). Aging effects and working memory in garden‐path recovery. Clinical Archives of Communication Disorders, 2(2), 91–102. [Google Scholar]
  89. Yoo, H. , McNeil, M. R. , Dickey, M. W. , & Terhorst, L. (2022). Linguistic and nonlinguistic processing speed across age‐matched normal healthy controls and individuals with left‐hemisphere damage, with and without aphasia. Aphasiology, 36(2), 147–169. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1 

TOPS-18-0-s001.pdf (100.7KB, pdf)

Articles from Topics in Cognitive Science are provided here courtesy of Wiley

RESOURCES