Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jul 23.
Published in final edited form as: Cortex. 2025 Jul 23;190:286–303. doi: 10.1016/j.cortex.2025.07.003

Abstract Word Dropout and Cross-Speaker Misalignment of Word Concreteness are Features of Conversation in Aging

Jamie Reilly 1,2, Virginia Ulichney 2, Benjamin Sacks 2, Anna Duncan 2, Sarah M Weinstein 3, Tania Giovannetti 2, Chelsea Helion 2, Gus Cooney 4
PMCID: PMC12327811  NIHMSID: NIHMS2099584  PMID: 40749579

Abstract

Language development is significant for an abstractness shift where the early concrete lexicon of childhood steadily expands to incorporate abstract concepts (e.g., mental states, emotions, ideas). Much remains to be learned about the trajectories of concrete versus abstract words across the span of development, especially within the domain of discourse production. We investigated the prevalence of abstract word use in dyadic conversations between adults of different ages (range 18-66). Our aim was to adjudicate between two opposing trajectories of language use in normal cognitive aging. The vocabulary growth perspective holds that aging is associated with a steady accrual of vocabulary knowledge resulting in greater lexical diversity and a larger pool of abstract words to draw upon in conversation. In contrast, resource pruning predicts the opposite trajectory such that age-associated gains in vocabulary acquisition are offset by diminished executive resources. Since abstract words tend to have higher lexical retrieval demands than concrete words, normal aging will result in an apparent dropout of abstract words. Moreover, this concreteness effect will be amplified within a resource-intensive communication modality such as conversation. We analyzed distributions of abstract and concrete words and cross-speaker alignment within unscripted conversations between adults of different ages (N=1565 conversations, >8 million words, age range 19-66). Aging was associated with abstract word dropout, and misalignment between conversation partners on concreteness expanded in parallel with their respective age differences. These results add to a growing body of research that implicates executive functioning in controlled lexical retrieval of abstract words. We discuss significance of these results for understanding intergenerational communication and relationships between aging, executive functioning, and lexical retrieval in high-level discourse.

Keywords: Conversation Analysis, Aging, Alignment, Natural Language Processing, Semantics

1. Introduction

Much of our understanding of language processing has been informed by paradigms such as lexical decision, speeded naming, priming, picture-word interference, single word semantic judgments, sentence grammaticality judgments, and dichotic listening (among others) (Alario et al., 2000; Balota et al., 2004; Dell’Acqua et al., 2010; Dhooge & Hartsuiker, 2011; Juhasz, 2005; Lupker, 1979; Pexman et al., 2016; Reilly et al., 2007; Schriefers et al., 1990, 1990; Yap & Balota, 2009). Each of these methods offers an advantage of experimental control at an expense to ecological validity. Real-world language use is rarely constrained to single words that come carefully matched across scores of psycholinguistic variables (Cutler, 1981). Moreover, the behavior of single words does not linearly scale to account for the dynamics of higher-level discourse. As such, there is a compelling need to study language in situ within the context of interactive production tasks such as dyadic conversation.

An ideal conversation is a cooperative endeavor where all participants calibrate the form and content of their own production to align with each other’s needs and abilities (Grice, 1975; Kita & Özyürek, 2003; Menenti et al., 2012; Pickering & Garrod, 2004; Xu & Reitter, 2015). This process of synchronization, also known as interactive alignment, spans numerous domains including: prosody, speech rate, emotional coloring, shared background knowledge, eye contact, body posture, and gesture (Ames et al., 2015; Rasenberg et al., 2020). Lexical-semantic complexity is one aspect of interactive alignment involving word choices. For example, the decision to produce ‘dog’ or ‘canine’ during a particular exchange is motivated by numerous factors, including vocabulary knowledge, social status, and prescriptive norms of formality dictated by the conversation setting (e.g., laboratory vs. a bar) (Grice, 1975).

One of the most widely investigated indices of lexical complexity in language is concreteness. Concrete words denote objects that can be seen, heard, touched, and experienced through the senses (Paivio, 1991; Crutch & Warrington, 2005; Paivio et al., 1994; Binder et al., 2005; Walker & Hulme, 1999), whereas abstract words reference ideas, emotions, and philosophical principles (e.g., rules, social conventions, scientific terms) that transcend the immediate sensorimotor present (Borghi, 2022; Crutch & Warrington, 2005; Reilly et al., 2016; Troche et al., 2014). One of the hallmarks of language maturation during late childhood is an ‘abstractness shift’ where expressive language expands to include a larger repertoire of abstract concepts ultimately becoming roughly balanced in frequency of lexical use (abstract ≈ concrete) in the adult English lexicon (Boulenger et al., 2007; Brysbaert et al., 2000; V. Coltheart et al., 1988; Kearney et al., 2025; Pinker, 1995; Ramey et al., 2013).

Concrete words tend to show persistent advantages relative to abstract words across numerous cognitive domains (e.g., word recognition, recall accuracy, age-of-acquisition, resilience to brain injury) (Franklin et al., 1994, 1994; M. Coltheart, 2000, 2006; Sandberg & Kiran, 2014). The majority of past research has attributed these processing advantages collectively termed the word concreteness effect to qualitative differences in representational frameworks and associative networks that support abstract and concrete concepts (Borghi et al., 2019; Crutch & Warrington, 2005; Kousta et al., 2011; Reilly et al., 2024; Vigliocco et al., 2014; Villani et al., 2021). However, recent work suggests that word concreteness effects cannot exclusively be attributed to semantic factors. Phonological and morphological complexity differences also mark abstract and concrete words (Reilly & Kean, 2007; Reilly et al., 2017; Hernandez et al., 2021; Kearney et al., 2025; Lievers et al., 2021).

One general point of consensus is that abstract words are harder than concrete words in terms of the executive resource demands required for retrieval and processing (Gleitman et al., 2021; Noppeney & Price, 2004). Hoffman and colleagues have advanced perhaps the most compelling and theoretically grounded perspective on the relation between executive dysfunction and abstract word retrieval, with foci on apparent dissociations (i.e., reverse concreteness effects) in acquired neurological disorders (Hoffman et al., 2011; Hoffman, Jones, et al., 2013; Hoffman et al., 2014; Hoffman & Lambon Ralph, 2011) and in normal aging (Hoffman et al., 2010, 2018). A common theme across this research is that processing demands of abstract words have numerous roots extending beyond qualitative differences in single word meaning to the broader structures (e.g., sparsity, diversity) of the associative lexical networks that support abstract vs. concrete words in context (see also Crutch & Warrington, 2005; Hoffman, Lambon Ralph, et al., 2013).

1.1. Aims of the Current Work

Our aim was to investigate the prevalence of abstract and concrete word use in spontaneous unscripted conversations between adults spanning a wide age range from late adolescence to early retirement (18 to 66 years). We reasoned that normal cognitive aging represents an in vivo model for a gradual reduction in executive functioning (for largescale empirical evidence see Hartshorne & Germine, 2015). These slowly evolving reductions in processing resources also occur in the context of a continuous accrual of vocabulary knowledge. The average 20-year-old speaker of American English knows about 42,000 words and will accrue thousands more lemmas by the age of 60 (Brysbaert et al., 2016) potentially conferring a more expansive lexical repertoire to draw upon in discourse production. This combination of vocabulary augmentation and executive depletion leads to two paradoxically opposing predictions, we term vocabulary growth vs. executive resource pruning.

The vocabulary growth perspective holds that people steadily accrue word knowledge over the lifespan (Brysbaert et al., 2016; Kavé, 2024). New concepts are constantly entering the lexicon (e.g., Snapchat, catfishing). One possibility is that later-life (i.e., post adolescence) lexical acquisition is roughly balanced across abstract and concrete concepts. If this indeed the case, the lexicons of older adults would be more extensive but not proportionately more abstract than younger adults. An entirely different possibility, however, is that lexical growth later in life is more heavily weighted toward abstract concepts learned from exposure to print and news (e.g., political and legal terminology, scientific and technological principles). Adult exposure to workplace and news-related language would theoretically shift the balance of the lexicon toward a more abstract state as we age, although these trajectories would be moderated by individual differences and cohort effects (e.g., profession, interests, language environment).

It is reasonably uncontroversial that people acquire many more words as they age. The questions of which words are acquired and how lexical acquisition is moderated by cohort effects (e.g., societal changes in technology use) persist. The vocabulary growth position supports the prediction that aging will confer a more expansive lexicon and that older adults will in turn have a larger repertoire of abstract words to draw upon relative to their younger peers. This prediction is predicated upon two major assumptions. First, a larger vocabulary acquired over the span of adulthood will incorporate more abstract words by sheer volume regardless of its balance (abstract > concrete). Second, more extensive vocabulary knowledge accrued over the lifespan will translate to more diverse and lexically-rich spontaneous language production.

Resource pruning, in contrast, supports the opposite prediction of vocabulary growth. Late adulthood (>age 50) is associated with accelerated reductions in processing speed, working memory, and inhibitory control (Higby et al., 2019). Regardless of passive vocabulary knowledge, resource pruning predicts that higher retrieval demands for abstract words will result in apparent abstract word dropout (or concrete word dominance). This concreteness bias will be especially evident in the context of an executively demanding communication modality such as conversation. Resource pruning predicts concrete word dominance in spontaneous discourse as a function of normal cognitive aging.

1.2. Executive demands and concreteness effects in conversation

Conversation is an executively demanding task especially when interlocutors are not well-known or are perceived as dissimilar to us (e.g., age, social status, education level, political ideology). Topic maintenance requires facile decisions about engagement (e.g., when to engage, who to engage with, how deeply to engage). These actions require us to make numerous inferences about our partner’s mental state (e.g., understanding of situation model, fatigue) while simultaneously monitoring our own production. A skilled conversationalist avoids behaviors such as monopolizing, repetition, and off-topic verbosity that would unbalance the reciprocal social contract of an interaction. These self- and other-monitoring skills require working memory, inhibitory control, cognitive flexibility, attentional vigilance, and theory of mind to achieve and preserve alignment.

The executive demands that scaffold alignment are also likely to differ by alignment modality (e.g., speech, gesture, writing) and level (e.g., lexical, affective). One useful distinction between levels of alignment was recently proposed by Gandolfi and colleagues (2023) in the distinction between situation model alignment and linguistic alignment. Situation model alignment involves establishing and maintaining common ground (i.e., shared understanding of a particular topic) between partners. For example, people of a similar chronological age have often endured similar historical circumstances (e.g., wars, recessions, popular culture trends) that confer some degree of communal common ground. As such, age can be a helpful factor in promoting situation model alignment when conversing about history. In contrast, linguistic alignment primarily involves modulation of form (e.g., lexical complexity) to accommodate the interlocutor’s capacities and expectations (e.g., modulating the complexity of a topic to explain to a child vs. an adult colleague) (Menenti et al., 2012; Pickering & Garrod, 2004). Abstract word use in conversation may play a significant role in linguistic alignment in terms of synchronizing lexical and semantic complexity.

Mazzuca and colleagues (2025) hypothesized that abstract words tend to facilitate conversations in several different ways. First, people tend to prefer starting conversations by referencing social and emotional states (e.g., ‘How are you feeling?’) that require the use of abstract words (Fini et al., 2021, 2023; Gandolfi et al., 2023). In contrast, a higher prevalence of concrete words may signal willingness to cede a turn or end the conversation (Mazzuca et al., 2025). In addition to social salience, the meanings of abstract words also tend to be indeterminant relative to concrete words. For example, an abstract statement (e.g., ‘Justice means a lot to me’ ) invokes uncertainty that begs one’s communication partner for elaboration.. These ambiguity must be reconciled as a function of maintaining common ground (Heller & Brown-Schmidt, 2023; Horton, 2005; Stalnaker, 2002). In contrast, a concrete statement such as ‘I need a new desk’ is more determined and less likely to require elaboration. Sense ambiguity as a feature of abstract words is further supported by Zradizolva, Sidhu, and Pexman (2018) who examined the content of language and gestures required by partners in guessing the meanings of abstract and concrete target words in a ‘taboo’ game format where they were not permitted to use the target word itself. Study participants tended to describe abstract target words using language that denoted people and introspections (i.e., observations about mental processes) and metaphorical gestures, whereas concrete targets were described using entity and object terms (e.g., sensorimotor attributes) and iconic gestures.

Past studies of concreteness effects in discourse have relied on simulation paradigms such as asking participants to continue an imagined conversation that starts with an abstract phrase or to evaluate the concreteness of a phrase that might occur in a conversational context (Fini et al., 2021; Mazzuca et al., 2025; Villani et al., 2022). We know of no past research that has examined the distribution of concreteness in unscripted conversations as a function of cognitive aging. In the section to follow, we describe how aging represents a viable ‘lesion model’ for executive dysfunction as manifested in conversations.

1.3. Brain aging as an in vivo model of executive decline

Human brain aging is marked by changes in anatomy and physiology that vary by rate, magnitude and localization across numerous interacting scales (e.g., molecular, cellular, genetic, vascular). At a microstructural level, aging impacts cell senescence, synaptogenesis, dendritic density, and vascular integrity (Erickson et al., 2012; Sikora et al., 2021; Tapia-Arancibia et al., 2008; Vinke et al., 2018). Many of these alterations in neural microstructure manifest as macroscale features (e.g., gyral/sulcal volume, functional connectivity) that can be quantified using neuroimaging methods. Much of our understanding of age-associated cognitive phenotypes (e.g., emergence of word finding deficits) is based on regional patterns of cortical thinning, functional reorganization, and connectivity between networks that serve language and cognition.

There evidence base linking aging to diminished executive functioning, processing speed, and lexical retrieval in behavioral tasks is extensive (Goral et al., 2007; Higby et al., 2019; Wierenga et al., 2008). Less clear, however, is the extent to which emerging deficits in word retrieval reflect compromise of language and/or semantic networks relative to a supporting cast of attention and multiple demand networks (Hoffman & Morcom, 2018; Shafto & Tyler, 2014; Zhang & Diaz, 2023). The field has faced several challenges in answering this question, including spatial overlap between adjacent networks and consensus regarding which brain regions comprise language and semantic networks relative to more domain-general cognitive control networks (Blank et al., 2014; Fedorenko, 2014; Fedorenko et al., 2012, 2013, 2013; Fernandino et al., 2022).

Vinke et al (2018) examined macrostructural changes in healthy aging within the Rotterdam cohort (>10k MRI scans) and reported non-linear tissue loss that accelerated after about the age of 50 (white matter > gray matter, males > females). The most extensive tissue thinning across the cohort occurred within the frontal lobes (adjusted for total intracranial volume) with the exception of older men who showed temporal lobe atrophy. Age-related atrophy within frontal structures such as reported by Vinke et al (2018) has long been associated with emerging deficits in inhibitory control (Cabeza & Dennis, 2012; Moscovitch & Winocur, 1995). People typically show some degree of compensation for tissue thinning through functional reorganization and network rewiring. Some of the major changes in function with advancing age include dedifferentiation (i.e., lower modularity, more diffuse activation patterns) a posterior-to-anterior shift (PASA), and right hemisphere recruitment under high task demands. These patterns have been described as either maladaptive or compensatory by models such CRUNCH (Compensation-Related Utilization of Neural Circuits Hypothesis) (Festini et al., 2018; Park & Reuter-Lorenz, 2009; Reuter-Lorenz & Cappell, 2008) and HAROLD (Hemispheric Asymmetry Reduction in Older Adults) (Cabeza, 2002) and more recently by ‘network disconnection’ hypotheses (Fjell et al., 2017).

1.4. Phenotypic expression of age-based executive deficits in lexical retrieval

Normal cognitive aging does not typically harm communicative functioning to the point where people experience social and/or occupational impairment. Yet, lexical retrieval deficits such as proper noun anomia (i.e., retrieving the names of acquaintances and public figures) are among the most common cognitive complaints of older adults (Goral et al., 2007; Higby et al., 2019). An important distinction for the current study is that age-related lexical retrieval difficulties are not typically rooted in degraded semantic knowledge (Wierenga et al., 2008). Retrieval failures in normal aging are often accompanied by tip-of-the-tongue states, circumlocutions, and self-cuing behaviors that indicate grossly intact object and/or person knowledge (Cohen & Faulkner, 1986; Maylor, 1990). These repair strategies in aging differ in kind and severity from patterns of anomia incurred in the context of multimodal semantic impairments in Alzheimer’s Disease and Primary Progressive Aphasia (Garrard, Lambon Ralph, et al., 2005; Reilly et al., 2011; Woollams et al., 2008).

Anomia and retrieval slowing at the single word level compound to impact higher level discourse. Older adult discourse is often stereotyped but has also been formally assessed as slower, more repetitive, circumlocutory, and off-topic relative younger adults (Dennis & Hess, 2016; Goral et al., 2007; Horton et al., 2010; Kavé & Goral, 2017; Williams & Nussbaum, 2013). Our key assumption is that at a population level these characteristics of discourse in aging are driven by resource limitations (e.g., inhibitory control, processing speed, working memory) rather than degraded semantic knowledge.

1.5. Study Aims

We examined the distribution of word concreteness in unscripted conversations using concreteness, aging, and conversation as coarse, quasi-experimental (or naturalistic) manipulations of executive demands. We pursued a big data approach by analyzing the CANDOR corpus (Reece et al., 2023), one of the largest and most extensively annotated conversation corpora currently available. CANDOR contains unscripted dyadic transcripts along with associated meta-data from over 1500 conversations (>8 million tokens).

We analyzed the CANDOR conversation transcripts using an open-source R software package we have termed ConversationAlign (Reilly et al., 2023). ConversationAlign is capable of importing, formatting, and cleaning raw conversation transcripts (e.g., omitting stopwords, lemmatizing, vectorizing) and appending published lexical norms for up to 30 user-specified dimensions (e.g., concreteness, length, valence) to each content word in the transcript. ConversationAlign accordingly converts raw language data (i.e., words) into quantitative time series aggregated by turn, interlocutor, and conversation. During the final stages of the processing algorithm, ConversationAlign computes main effects of each target dimension and two complementary indices of linguistic alignment between interlocutors (e.g., turn-by-turn covariance between Person A and Person B on word concreteness).

We processed conversation transcripts with a focus on main effects and linguistic alignment for word concreteness as a function of aging relative to five additional comparison dimensions (i.e., valence, trust, word frequency, word length, arousal). Our aim was to evaluate two contrasting outcomes predicted by vocabulary growth vs. executive resource pruning perspectives on lexical retrieval in aging (see Section 1). Vocabulary growth predicts higher prevalence of abstract words in conversation with advancing age secondary to more extensive crystallized vocabulary knowledge. In contrast, the prediction of executive resource pruning is that aging will bias conversation toward concrete secondary to higher retrieval demands for abstract words.

In addition to main effects, we examined linguistic alignment as functions of age differences between conversation partners. We predicted that increasing age gaps between conversation partners would result in higher levels of mismatch or misalignment in abstract/concrete word use. That is, if older adults show abstract word dominance, conversations will be misaligned with greater abstract than concrete word use between the older and younger conversation partner. In contrast, if older adults experience abstract word dropout, conversations will be misaligned due to greater concrete than abstract word use between the older and younger conversation partner.

2.0. METHOD

2.1. Transparency and Openness

Given shifts from our pre-registration, and in particular, updates to the specific analyses used and our focus primarily on concreteness, this work may be considered exploratory. We preregistered hypotheses and analysis plans (visit at https://osf.io/ft9h7) and catalogued deviations in the appended supplemental materials section according to recent guidelines proposed by Willroth and Atherton (2024). Notably, we primarily focus on concreteness here (see introduction for rationale), but also included all of the original pre-registered variables for comparison. Detailed analyses examining pre-registered dimensions other than concreteness are included in supplemental materials. All conversation data are publicly available via the Candor corpus (Reece et al, 2023). The ConversationAlign R package is open-source and freely available for inspection and use. Associated lookup databases and associated code are available on the OSF at: https://osf.io/atf5q/?view_only=3abcf674d398421eb2d4f9d3e88e6063. The ConversationAlign R package can be downloaded and installed via Github at https://github.com/Reilly-ConceptsCognitionLab/ConversationAlign.

2.2. Ethics and Human Subjects Protections

We conducted analyzed de-identified conversation transcripts from the CANDOR corpus (Reece et al, 2023). No new human subjects were enrolled for this study. Details of human subjects protections and data dissemination restrictions can be found in Reece et al. (2023).

2.3. Participants and Conversation Corpus

We extracted and analyzed transcripts (N=1565) with at least a one year age-difference between interlocutors based on self-reported age (not date of birth). CANDOR contains unscripted conversations spanning over eight million words of words of spoken American English from a diverse sample (e.g., age, race, gender). Figure 1 and Table 1 outline participant demographics. Age of the participants ranged from 19 to 66 (mean=33.1 yrs) with a moderate positive skew (skew=.81) toward younger (see Figure 1, Panel A). The average age difference between conversation partners was 13.1 years (range 1-44) (see Figure 1, Panel B). Many conversations occurred between partners of different genders (52%) and races/ethnicities (55%). Participants on average completed 15.7 years of formal schooling (SD=1.95 yrs) with an average education difference between interlocutors of 2.1 years (SD=1.81 years). Table 2 summarizes conversation corpus analytics.

Figure 1. Participant Demographics and Corpus Counts.

Figure 1.

Open in a new tab

Table 1. Participant Demographics.

Demographic/Factor (N = 1359) Mean/Total SD Range
Age (mean) 33.1 years 11.13 19 to 66
Education (mean) 15.68 years 1.95 10 to 21
Gender (self-reported) Male 581 (43%)
Female 752 (56%)
Other/NR 26 (1%)
Race/ethnicity (self-reported) Asian 192 (14%)
Black/AA 114 (8%)
Hispanic/Latino 99 (7%)
White 878 (65%)
Other/NR 5%
Age difference (avg per conversation) Cleaned 13.1 9.62 1-44
Education difference (in years) Cleaned 2.11 1.81 0 to 11
Open in a new tab

Table 2. Conversation Corpus Analytics.

Conversation corpus analytics (N = 1565) Mean/Total SD Range
Total word count (all conversations) Raw 8,701,010 n/a n/a
Cleaned 4,525,572 n/a n/a
Word count (mean per conversation) Raw 5559.75 1509 2005-16816
Cleaned 2891.73 779 996-9153
Exchange count (mean per conversation) Cleaned 440.91 156.70 132-1779
Words per-turn (mean per turn) Cleaned 3.52 1.13 1.08-12.72
Age difference (avg per conversation) Cleaned 13.1 9.62 1-44
Education difference (in years) Cleaned 2.11 1.81 0 to 11
Gender different? Cleaned 814 different (52%)
Race/ethnicity different? Cleaned 862 different (56%)
Open in a new tab

Note: We defined a turn as all uninterrupted words spoken by one person until the utterance was broken by their partner. An exchange was two successive turns.

2.4. Transforming Text to Time Series Data via ConversationAlign

We used ConversationAlign (Reilly et al., 2023) to process each of the raw conversation transcripts from the CANDOR corpus. This software transforms raw language transcripts into numeric time series using a combination of custom regular expressions and wrapper functions from other natural language processing packages such as stringi (Gagolewski & Tartanus, 2023), textclean (Rinker, 2018b) and textstem (Rinker, 2018a). For a more detailed treatment of the software code and algorithms that constitute ConversationAlign, see Sacks et al (2025).1

During cleaning, ConversationAlign transformed text to lowercase, deleted non-alphabetic characters (e.g., numbers, symbols, line breaks, unicode symbols), expanded contractions (e.g., it’s → it is) and omitted stopwords. We originally omitted stopwords by indexing a customized stopword list that was generated by first extracting all single words from the MIT Media Lab stopword list (Bertsimas, 2017).2

We then appended multiword utterances (N=583) that corresponded to idioms, greetings, and collocatives (e.g., raining cats and dogs, red tape, Merry Christmas). For transparency, we executed all of the original analyses using this original (MIT) stopword list. Upon inspection of the MIT list during peer-review, it was pointed out that this list contained numerous open-class words (e.g., ask). This unexpected observation led us to formally analyze the MIT stopword list with the goal of identifying and omitting any spurious open-class words. We first marked each word in the original MIT stopword list with its corresponding universal part-of-speech (UPOS) code using the ‘udpipe’ package (Wijffels, 2023). We then filtered the tagged stopword list, retaining determiners (e.g., the, a), adpositions (on, in), filler words (e.g., .uh, um), pronouns (e.g., he, she, the), numerals (0-12), modal auxiliaries (e.g., should), idioms (e.g., raining cats and dogs) and greetings (e.g., Hello, How are you?) .3 We then re-executed all of the original analyses using this revised, more conservative stopword list. The original version of the manuscript reporting liberal stopword omissions (using MIT stopwords) is archived and can be inspected on the OSF at (https://osf.io/atf5q/).

Following text cleaning and stopword omission, ConversationAlign lemmatized each word using the textstem R package (Rinker, 2018a), converting morphological derivatives to their corresponding dictionary entries. During the final formatting step, ConversationAlign split the cleaned and lemmatized text into a one-word per-row format. This process yielded a vector of content words marked by turn, speaker, conversation, and any other user-specified metadata. For analysis purposes, ConversationAlign considers a turn as all uninterrupted tokens produced by one interlocutor until the second interlocutor produced at least one token word (see Reece et al., 2023). An exchange is two consecutive turns between interlocutors. For the analyses we report to follow, we yoked values for concreteness (Brysbaert et al., 2014), trust (Raji & da Melo, 2020), word length (in letters), log10 word frequency (Brysbaert & New, 2009), valence (Mohammad, 2018) and arousal (Mohammad, 2018), to each content word in the CANDOR transcripts. Table 3 details published sources and scales of each dimension.

Table 3.

Psycholinguistic variables for computing alignment

Descriptive Statistics by Lexical Dimension and Age Band

Variable Description Scale
Concreteness Extent to which a word can be experienced through sensory modalities (e.g., seen, heard, touched) 0-10 rescaled
Arousal Extent to which a word evokes physiological arousal 0-10 rescaled
Valence Unpleasant to pleasant 0-10 rescaled
Trust Semantic distance from each target word to the word, ‘trust’ as assessed through word embeddings 0-10 rescaled
Length Total number of letters in each word Raw (unscaled)
Word Frequency Word frequency in English Log10 per million Raw (unscaled)
Note: Concreteness (Brysbaert et al., 2014); Arousal (Mohammad, 2018); Valence (Mohammad, 2018); Trust (Raji & da Melo, 2020); Word Frequency (Brysbaert & New, 2009)
Age Band
Variable 19-29 30-39 40-49 50-59 >60 Δ%
Concreteness 3.89 (.22) 4.01 (.23) 4.10 (.23) 4.13 (.23) 4.21 (.28) 8.2%
Arousal 4.13 (.13) 4.15 (.12) 4.16 (.11) 4.18 (.12) 4.18 (.11) 1.2%
Frequency 4.48 (.10) 4.45 (.10) 4.46 (.10) 4.43 (.09) 4.43 (.10) −1.1%
Word length 4.30 (.17) 4.32 (.18) 4.30 (.17) 4.31 (.18) 4.31 (.16) 0.1%
Valence 6.49 (.17) 6.43 (.17) 6.39 (.17) 6.38 (.16) 6.33 (.19) −2.5%
Trust 3.43 (.06) 3.41 (.06) 3.40 (.06) 3.38 (.05) 3.38 (.06) −1.4%
Note: Δ% reflects change from Band 1 to Band 5 as derived by: (final-initial)/∣initial∣. Statistics reflect mean and SD of each dimension.
Open in a new tab

2.5. Alignment Indices

ConversationAlign computes two indices of alignment for each conversation. The Spearman R-value quantifies turn-by-turn covariance between two interlocutors on a given dimension. The Area Under the Curve of the difference (dAUC) between interlocutors yields a normalized index for how globally aligned two partners were on any given dimension (e.g., concreteness, valence, word length) across a conversation. Figure 2 represents an overview of what the different alignment indices and main effects mean in the context of a hypothetical conversation between Mary and Dave. Perhaps Dave has experienced a traumatic brain injury and is undergoing a metacognitive communication intervention designed to augment both his baseline pleasantness and his capacity to align with the pleasantness of a conversation partner. Dave’s psychologist has identified valence (or pleasantness) as the lexical variable of interest in his exchange with Mary.

Figure 2. ConversationAlign: Interpreting Alignment Indices.

Figure 2.

Open in a new tab

Note: The AUC between interlocutors is isomorphic with their difference time series, hence the abbreviation dAUC as illistrated above.

Figure 2 illustrates several ways in which Dave and Mary might be aligned/misaligned in valence over the span of their conversation. The simplest index of alignment is a main effect of valence (i.e., Dave > Mary) averaged across all the content words of their conversation (see Panel 4). Another dimension of alignment as illustrated in Panel 5 occurs at the turn level as measured by a simple Spearman correlation. For example, if Dave immediately and consistently produced unpleasant (i.e., low valence) responses in lockstep with Mary, this behavior would yield a high positive correlation. A high positive Spearman R value at a lag of 0 would reflect immediate synchrony at the turn level (e.g., I say something mean, you respond with something mean) (for discussion of alignment approaches see Xu & Reitter, 2015).

ConversationAlign computes a third complementary measure of global alignment (dAUC) between interlocutors that is less sensitive to turn-by-turn variability and instead reflects slower changing differences. This dAUC index is illustrated in Figure 2, Panel 5 in the context of our hypothetical interaction between Dave and Mary. In this particular exchange, Dave has a lower baseline valence than Mary and although he begins to close this gap across the conversation, it is always apparent. This global difference can be measured by calculating a normalized dAUC between interlocutors. The y-axis when determining the dAUC value reflects the scale of the variable of interest (e.g., valence 0-9), whereas the x-axis represents time (as assessed by all successive turns of a conversation). At turn 1, the valence difference between Mary and Dave reflects the absolute value of the difference between the averaged valence values of all words produced by each speaker at that turn. We derived a difference AUC for each conversation by summating the differences across turns and applying the trapezoidal rule to determine dAUC. Lower dAUC values indicate closer global alignment on any particular variable of interest.

Spearman R is a unitless measure that quantifies turn-by-turn correspondence between two interlocutors. In contrast, dAUC is proportional to conversation length. Longer conversations will typically inflate the dAUC between interlocutors regardless of meaningful differences on a given target dimension. Therefore, standardization of conversation length (i.e., turn count) is necessary to compare conversations of different lengths. We standardized turn length by down-sampling all conversations to 133 exchanges (i.e., the briefest conversation between two speakers of different ages by total number of exchanges in the CANDOR corpus).

2.6. Statistical Analyses

We report descriptive statistics and non-parametric correlations (Spearman rho) quantifying bivariate relations between demographic (e.g., age, age differences) and discourse level factors (e.g., word count) across all conversation transcripts. We examined linear changes in semantic, affective, and lexical dimensions of conversational word use with a focus on age-related variability in the prevalence of concreteness, arousal, valence, word frequency, word length, and trust. We examined variability in alignment as a continuous function of age differences between conversation partners.

We then tested whether indices of alignment strength would shift in tandem with age difference between speakers. Specifically, we utilized the dAUC metric of alignment for relevant linguistic dimensions as dependent measures in a multi-membership regression model executed using the “lmer” function from the R “lmerMultiMembership” package (van Paridon et al., 2023). Notably, according to Reece & colleagues’ (2023) original article, in the full CANDOR dataset, half of the sample were involved in at least two conversations; this means that in data examined here, some participants appear in multiple conversations, which we account for by using a multi-membership model. In this analysis, we examined how age differences between speakers in discussions were associated with concreteness alignment between speakers (see supplemental materials for additional analyses). The model structure was as follows: dAUC ~ age_difference + gender_diff + education_diff + race_diff + (1∣Participant_ID) + (1∣Dyad). We included age difference (absolute value) between conversation partners as our primary fixed effect of interest, and found that a model also including speaker age as a covariate does not fit the data better according to a likelihood ratio test run with the “anova()” function from the “stats” R package (R Core Team and contributors, 2024). Covariates included difference in self-reported gender (same or different), education difference (absolute value of the difference in approximate years of reported education), and race/ethnicity difference (same or different) between conversation partners, and we include random intercepts for individual and dyad (see supplemental materials for additional models). We also include models examining individuals’ mean usages (as opposed to dyadic alignment) of each linguistic dimension in conversation associated with these same predictors and covariates as well as speaker ages , and with random effects of individual and dyad, in supplemental materials.

3. RESULTS

3.1. Conversation Analytics as a Functions of Aging

We analyzed 4.5 million lemmatized words of spoken dialogue from unscripted conversations produced by 1359 speakers of American English. Figure 3 (Panel A) illustrates bivariate relations between demographic, psycholinguistic, and discourse characteristics.

Figure 3. Bivariate Relationships: Demographics and Lexical Variables.

Figure 3.

Open in a new tab

Note: Variable key: concrete: correlation of concreteness values; edu: correlation years of education recoded to numeric from participant categorical self-report (e.g., high school=12 years formal education); n_turns: correlation reflecting number of turns per conversation; wc_raw: correlation with total number of words per conversation prior to cleaning and stopword removal; wc_clean: correlation with total number of words per conversation after cleaning and stopword removal; wc_turn: correlation with the average turn length (i.e., strings-per-turn) after cleaning; ContentRatio: correlation with the ratio of wc_clean/wc_raw reflecting the prevalence of open class words in the conversation; TTR: correlation with Type Token Ratio reflecting the ratio of unique words to total number of words in each cleaned conversation; Blank cells reflect non-statistically significant correlations. Individual correlations were statistically significant at a Bonferroni-corrected p-value < .001 (56 conttrasts). We derived baseline change scores for each lexical variable in Panel B by subtracting each person’s average value for that particular variable across all content words relative to the average produced by all 19-20 year olds in the Candor Corpus (N=143).

Aging was associated with higher output of words-per-conversation both in terms of raw word count (Spearman R =0.25) and word counts for cleaned/lemmatized conversations (Spearman R=0.18). The ratio of cleaned relative to raw word count yields a coarse index of semantic content per conversation (i.e., raw word count minus stopword count = clean word count). Aging was strongly negatively associated with content ratio (Spearman R= -0.49), indicating that older adults tended to produce more words but with greater semantic sparsity than young adults as reflected by relatively higher stopword ratios. One additional demographic association was a weak but statistically significant positive correlation between age and years of education (Spearman R=0.27). The age distribution of the sample was skewed toward younger adults, many of whom may have still been completing school. As such, older adults tended to have more years of formal education, but this association between age and education in the CANDOR sample was small.

Table 1 summarizes the distribution of each lexical dimension by decade (see also Figure 2B for continuous plot).4 Concreteness showed the strongest age effects characterized by a 8.2% rise (Spearman R=.41) across adulthood (age 19 to 66). Although several other lexical variables also showed statistically significant associations with age, the practical significance of these language changes were nominal (<2.5% all). Aging was associated with weak negative correlations between lexical distributions for trust (Spearman R=.31, % change =−1.4%), valence (Spearman R=30, % change =−2.5%), and word frequency (Spearman R=.16, % change =−1.1%).

3.2. Alignment of Word Concreteness ~ Age and Age Differences

Figure 4 illustrates turn-by-turn (Spearman rho) and global (dAUC) alignment dynamics for word concreteness across conversation partners. For details of alignment and model summaries across the remaining comparison dimensions (e.g., word length, frequency), see supplemental materials. Alignment on concreteness as indexed by dAUC ranged from a maximum of 267.74 (least aligned) to 32.08 (most aligned) with a mean dAUC of 126.08 and a standard deviation of 48.57. Figure 4 illustrates bivariate correlations between dAUC and age difference between interlocutors (Panel B, R=.13) and age of the individual (Panel C, R=.20).

Figure 4. Global Alignment Indices.

Figure 4.

Open in a new tab

Note: Panel A represents the average turn-by-turn alignment as indexed by the Pearson Correlation between partners across all conversations. Lag/lead represents turn-by-turn covariance offset by n-turns (e.g., Mary says something rude to Dave, and Dave tends to reliably reciprocate but only after two-turn delay. Panels B and C represent relationships between global alignment on word concreteness and age difference (Panel B) and baseline age (Panel C) across conversation partners.

Table 4 reflects turn-by-turn linguistic alignment on concreteness at five different turn lags as indexed by Spearman rho correlations. For example, perfect mirroring of concreteness (i.e., I say something abstract, your reply is equally abstract) would be evident in a strong positive correlation at a turn lag of zero. Figure 4 (Panel A) represents variability in turn-by-turn covariance between conversation partners on word concreteness.

Table 4. Turn-by-Turn Alignment on Concreteness as Indexed by Lagged Pearson R.

Rho Descriptive Turn Lag
−2 −1 0 1 2
Mean −0.02 −0.03 −0.03 −0.02 −0.01
SD 0.07 0.08 0.08 0.07 0.06
Min −0.29 −0.31 −0.27 −0.29 −0.21
Max 0.43 0.59 0.37 0.27 0.19
Open in a new tab

Note: Spearman rho values index alignment across concreteness values averaged across all content words aggregated by turn and speaker. A lag of 0 indicates an immediate response (e.g., Mary produces a concrete utterance; Dave immediately responds with a concrete utterance. Lags indicate an n-turn delay in alignment across conversation partners (e.g., Mary produces a concrete utterance; two turns later Dave produces a concrete utterance consistent with temporally delayed alignment).

Group level means across all lag/lead alignment conditions clustered near zero. These results suggest that conversation partners do not tend to align with one another on concreteness at the turn level. However, an exclusive consideration of means may likely obscure meaningful individual differences reflected in the range of the data at different lags. For example, some participants strongly mirrored one another on word concreteness producing correlations as strong as .59 (at lag −1 turn). That is, when Mary produced an abstract word, Dave tended to respond with an abstract word. Other participants showed paradoxical anti-correlations as low as −0.31 (e.g. when Mary produced an abstract word, Dave produced a concrete word).

Since there was no evidence for a group-level alignment pattern at the turn level (by correlation), we concentrated the final analyses on more global differences between conversation partners as indexed by dAUC. As outlined in the Analysis section, we fit a multi-membership model with dAUC as the outcome, with predictors including absolute age difference between speakers (in years) for each conversation. We also adjusted for the following covariates describing demographic variability between conversation partners (i.e., whether there were differences in reported gender, race/ethnicity, and approximate years of education) as well as random intercepts for individual and dyad, accounting for membership of some individuals to multiple dyads via multi-membership model.

We found that as the absolute age difference between interlocutors in conversation increased, they tended to grow more misaligned on word concreteness, b = 0.52, 95% CI [0.43, 0.61], SE = 0.05, t(3120) = 11.14, p < .001, R2 marginal = 0.08, R2 conditional > 0.99. For full results see Table 6.5 Including speaker age as an additional covariate in this model (AIC = −14,204, BIC = −14,149) did not fit the data better than our model of interest (AIC = −15,905, BIC = −15,857), X2(1) = 0.00, p > .99; this model of interest also did not fit the data substantially better than one only including random effects of individual and dyad (AIC = −16,890, BIC = −16,866), X2(4) = 0.00, p > .99.

Table 6. Regression table for dAUC linguistic concreteness multi-membership model across conversations.

dAUC, concreteness
Predictors Estimate
[95% CI] 		SE t p df
Intercept 171.78
[169.23, 174.33]		 1.30 131.82 <0.001 3120
Age difference (years) 0.52
[0.43, 0.61]		 0.05 11.14 <0.001 3120
Gender difference (ref. = same gender) 1.64
[0.24, 3.05]		 0.72 2.29 0.022 3120
Education difference (approx. years) 0.18
[−0.30, 0.66]		 0.24 0.74 0.457 3120
Race/ethnicity difference (ref. = same race) −1.53
[−3.35, 0.29]		 0.93 −1.65 0.099 3120
Observations 3130
R2 marginal / R2 conditional 0.081 / 0.999
Open in a new tab

Note: 95% CI = 95% confidence interval estimated via Wald method.

4.0. General Discussion

We investigated patterns of language use and linguistic alignment with a focus on word concreteness in unscripted conversations between people of different ages spanning adulthood. We reasoned that aging represents a viable ‘lesion model’ for changes in brain structure and function that have outsized impacts on executive functioning. Slowly progressive and non-linear reductions in inhibitory control, processing speed, and lexical retrieval after about the fifth decade of life may have substantial impacts on language even in healthy aging (Dennis & Hess, 2016; Goral et al., 2007; Kemper et al., 2001). Nevertheless, apart from global discourse characteristics such as slowing and dysfluency, we lack a granular understanding of how diminished lexical retrieval might impact conversation. Our goal was to adjudicate between two competing hypotheses generated from vocabulary growth vs. executive resource pruning perspectives on lexical retrieval in aging.

Relative to five other lexical control variables (e.g., length, valence), word concreteness was strongly associated with age. Each successive decade of life beyond the age of 30 was associated with steady change in the distribution of concreteness. These small differences by decade gradually compounded such that the lexical distribution of a 19 year-old was significantly more abstract than that of a 65 year-old. This trend toward abstract word dropout in aging is consistent with an executive pruning account of lexical retrieval. In addition, aging was not associated with higher lexical variability in conversation as indexed by a null correlation between age and type-token ratio (TTR). Instead, older adults tended to produce less content words per conversation as assessed by lower content word ratios (i.e,, word count clean/word count raw). Together these results suggest that aging was associated with a relative narrowing of lexical diversity characterized by more concrete language punctuated by higher rates of stopwords, fillers, and idioms.

In addition to main effects, we examined alignment as a continuous function of the age difference between interlocutors (range 1-44 years). Conversations grew increasingly misaligned by word concreteness with greater age discrepancies between conversation partners adjusted for numerous covariates (e.g., different race, different gender, different education level). The most robust main effect across all dimensions (older > younger) was concreteness relative to frequency, word length, valence, trust, and arousal. Younger speakers tend to use more abstract words in conversation, and advancing age was associated with of higher levels of concrete word use.

The observed trends may improve our understanding of the trajectory of language production across the span of normal adult development. We interpreted shifting patterns of word concreteness (i.e., slow but steady dropout of abstract words) in the context of generally preserved semantic processing as reflecting an executive contribution to lexical-semantic retrieval. Although lexical retrieval failures in aging are well-established, the etiology of word finding deficits remains debated. One specific point of contention is whether increased retrieval demands for abstract words are supported by an ‘all-purpose’ set of executive functions or instead whether word meaning is subserved by a more specialized semantic control mechanism (Badre & Wagner, 2002; Jackson, 2021; Jackson et al., 2021; Jefferies et al., 2008; Reilly et al., 2024). Although the data we reported here do not speak directly to questions of modularity in cognitive and/or semantic control, they do provide further evidence that the construction of word meaning is resource-demanding.

4.1. Strengths, limitations, and alternative accounts of the data

We pursued a ‘big data’ approach to conversation analysis across the adult lifespan, analyzing over four million words of conversation with a focus on concreteness. The sheer statistical power of this investigation is a strength that must be cautiously evaluated against a constellation of caveats. Threats to validity and generalization include aspects of the algorithm(s) we used to process the data and in demographic characteristics of Candor corpus (e.g., age skew, cultural homogeneity). We developed ConversationAlign as a tool for processing conversation transcripts. The software parses text into individual content words, appending psycholinguistic norms using a continuous bag-of-words (CBOW) approach (Liu, 2020). The primary limitation(s) of this method are its insensitivity to syntax, prosody, and pragmatic markers that imbue words in isolation with emergent meaning in connected discourse. Another source of error inherent to CBOW approaches involves lexical ambiguity. ConversationAlign matches content words in the target transcript to their corresponding entries in a lookup database. This procedure is agnostic to polysemy, part-of-speech, or other statistical/contextual cues that delineate bank as a river vs bank as a financial institution.

Another limitation of the current study involves inference. Our core assumption is that healthy aging prominently impacts executive functioning relative to other cognitive domains (e.g., semantic knowledge) and that changes in lexical retrieval are primarily attributable to executive resource pruning but not some other illusory ‘third’ variable (e.g., peripheral sensory deficits, motor programming). However, since we did not formally assess individual differences in executive functioning, the validity of our inference(s) rests on the key assumption that the observed effects are attributable to a canonical maturational process and not strategic and/or linguistic factors. For example, the tendency of older adults to use more concrete words in conversation reflects a cohort effect idiosyncratic to American English. Another possibility is that older adults concretize their language output as an accommodation strategy to connect with their younger counterparts. These potential confounds could be ruled out by analyzing conversational exchanges across other languages and cultures, or examining historical/archival conversation transcripts (Cheng et al., 2023).

In addition to alternate accounts of the data, another limitation of the current study is involves age distribution of the Candor sample which was skewed toward younger adults with a gradual winnowing of sample size into the mid-60s (max 66 years). Many researchers claim that age-related cognitive decline(s) only become prominent or accelerate after about the age of 55 (Rönnlund et al., 2005). In contrast, Salthouse (2009) has argued that subtle age-related deficits emerge decades earlier. In either case, the relatively sparse representation of older adults in the Candor sample is a limitation that must be addressed. Understanding the trajectory of lexical selection in conversation across the full spectrum of normal aging (ages 60 to 100) may be essential for discriminating healthy from pathological aging.

4.2. How Might Conversation and Concreteness inform Dementia Detection?

Our understanding of the trajectory of concreteness/abstractness in human development has largely been informed by two disparate bodies of evidence. Decades of early childhood language research supports the claim that concrete words are ‘first in’ relative to abstract words (Juhasz, 2005). An altogether different body of research has focused on explaining category-specific dissociations in dementia and aphasia. Recall that the word concreteness effect refers to the collective advantage people typically show for concrete relative to abstract words across many domains (e.g., word recognition, memory, age-of-acquisition). One of the more controversial phenomena in cognitive neuropsychology has involved case series of some patients with semantic dementia showing apparent reversals of the concreteness effect that are characterized by selective advantages for abstract over concrete words (Bonner et al., 2009; Breedin et al., 1994; Macoir, 2009; Papagno et al., 2009; Yi et al., 2007). Hoffman and Lambon Ralph (2011) argued that reverse concreteness effects are not in fact a typical feature of semantic dementia and may instead reflect artifacts of word frequency and how concreteness is measured in natural discourse.

It is reasonably well-accepted that people with dementia and/or primary progressive aphasia tend to produce repetitive, empty, and circumlocutory utterances (e.g., thing, stuff) (Garrard, Maloney, et al., 2005; Hodges et al., 1995; Hoffman et al., 2014; Hoffman, Jones, et al., 2013; Tomoeda et al., 1996). This language output is dominated by highly frequent words with relatively low information content, many of which would technically be considered stopwords in the current study. In real-life discourse, many stopwords (e.g., ‘thing’, ‘stuff’, ‘the’) would technically be classified as abstract via traditional rating scales where people are asked to, ‘rate the extent to which X can be experienced through the senses’ (for discussion see also Reilly et al., 2025). Consider, for example, the weak extent to which high frequency words such ‘thing’ or ‘the’ evoke multisensory experiences. The decision to factor concreteness of stopwords into a distributional analysis of discourse is likely to present a picture of a highly abstract language sample. Thus, the extent to which reversal of the concreteness effect represents a ‘real’ cognitive phenomenon versus a methodological artifact remains unclear.

In the current study, we did not evaluate people with dementia or probe semantic knowledge for abstract concepts. Nevertheless, the observed pattern of abstract word dropout in normal aging may provide benchmarks for detecting subtle lexical retrieval deficits in early stage dementia and in understanding amplified concreteness effects (e.g., abstract word anomia, abstract word deafness) in acquired aphasia (Franklin et al., 1994, 1995; Goral et al., 2007).

4.1. Concluding Remarks and Future Directions

One of the most exciting developments in language research over the past decade has involved modeling naturalistic discourse in situ. This paradigmatic shift has fueled the creation of new computational tools and methods such as ConversationAlign. We used this algorithm to analyze the distribution of word concreteness in conversations with the goal of adjudicating between two competing accounts of language processing in aging, vocabulary acquisition vs. executive resource pruning. A robust pattern of abstract word dropout is consistent with an executive resource pruning account (although see Section 4.1 for alternate explanations such as strategy use and cohort effects). Future efforts to confirm this executive locus of abstract word dropout in aging involve several promising options including:

  1. Analyzing lexical alignment dynamics in conversations in cultures and languages other than American English.

  2. Implementing controlled experimental procedures and manipulations (e.g., dual task interference) during conversations between older adults.

  3. Analyzing conversations across a wider spectrum of healthy aging (ages 60 to 100).

  4. Characterizing individual differences in processing speed, executive functioning, and language abilities using gold standard neuropsychological measures.

  5. Analyzing archival/historical conversations to rule out cohort effects.

Supplementary Material

1

Conflict of Interest and Funding Statement

We have no known conflicts of interest. All tools developed here are open source. This work was funded by a grant (R01 DC013063) from the National Institute on Deafness and Other Communication Disorders (NIH/NIDCD)

Footnotes

Prior Dissemination

See preprint on PsyArXiv (https://osf.io/preprints/psyarxiv/3crua)

1

ConversationAlign methods manuscript currently under review. Download preprint from PsyArXiv at https://osf.io/preprints/psyarxiv/7xqnp_v1

2

Inspect the original MIT stopword list at https://web.media.mit.edu/~lieber/Teaching/Common-Sense-Course/Stop-Words.Text. The original list newly tagged with part-of-speech codes and the revised stopword list (Temple_Stopwords25) are available for inspection and use on the OSF at .

3

An annotated walkthrough demonstrating these procedures for part-of-speech tagging and filtering can be viewed at https://reilly-lab.github.io/Jamie_Stopwords.html.

4

These R values are regression coefficients that are interpretable in the context of linear bivariate relationships. We tested for linearity of each relationship (e.g., arousal ~ aging, concreteness ~ aging) using rainbow tests of linearity within the ‘lmtest’ R package (Hothorn et al., 2022). The null hypothesis of a rainbow test is that a regression model is linear. See ‘lmtest’ vignette at https://search.r-project.org/CRAN/refmans/lmtest/html/raintest.html.

5

We generated Table 6 using the “sjPlot” R package (Lüdecke et al., 2024)

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Data Availability

All data and associated code are available via the Open Science Foundation at https://osf.io/atf5q/?view_only=3abcf674d398421eb2d4f9d3e88e6063

REFERENCES

  1. Alario FX, Segui J, & Ferrand L (2000). Semantic and associative priming in picture naming. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology, 53A(3), 741–764. [DOI] [PubMed] [Google Scholar]
  2. Ames DL, Honey CJ, Chow MA, Todorov A, & Hasson U (2015). Contextual alignment of cognitive and neural dynamics. Journal of Cognitive Neuroscience, 27(4), 655–664. 10.1162/jocn_a_00728 [DOI] [PubMed] [Google Scholar]
  3. Badre D, & Wagner AD (2002). Semantic retrieval, mnemonic control, and prefrontal cortex. Behavioral and Cognitive Neuroscience Reviews, 1(3), 206–218. 10.1177/1534582302001003002 [DOI] [PubMed] [Google Scholar]
  4. Balota DA, Cortese MJ, Sergent-Marshall SD, Spieler DH, & Yap MJ (2004). Visual Word Recognition of Single-Syllable Words. Journal of Experimental Psychology: General, 133(2), 283–316. [DOI] [PubMed] [Google Scholar]
  5. Bertsimas D. (2017). MIT Media Lab Stoword List [Computer software]. https://web.media.mit.edu/~lieber/Teaching/Common-Sense-Course/Stop-Words.Text [Google Scholar]
  6. Binder JR, Westbury C, & McKiernan K (2005). Distinct brain systems for processing abstract and concrete concepts. Journal of Cognitive Neuroscience, 17(6), 905–917. [DOI] [PubMed] [Google Scholar]
  7. Blank I, Kanwisher N, & Fedorenko E (2014). A functional dissociation between language and multiple-demand systems revealed in patterns of BOLD signal fluctuations. Journal of Neurophysiology, 112(5), 1105–1118. 10.1152/jn.00884.2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bonner MF, Vesely L, Price C, Anderson C, Richmond L, Farag C, Avants B, & Grossman M (2009). Reversal of the concreteness effect in semantic dementia. Cognitive Neuropsychology, 26(6), 568–579. https://doi.org/919094196 [pii] 10.1080/02643290903512305 [doi] [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Borghi AM (2022). Concepts for which we need others more: The case of abstract concepts. Current Directions in Psychological Science, 31(3), 238–246. 10.1177/09637214221079625 [DOI] [Google Scholar]
  10. Borghi AM, Barca L, Binkofski F, Castelfranchi C, Pezzulo G, & Tummolini L (2019). Words as social tools: Language, sociality and inner grounding in abstract concepts. Physics of Life Reviews, 29, 120–153. [DOI] [PubMed] [Google Scholar]
  11. Boulenger V, Dacoppet N, Roy AC, Paulignan Y, Nazir TA, & Décoppet N (2007). Differential effects of age-of-acquisition for concrete nouns and action verbs: Evidence for partly distinct representations? Cognition, 103(1), 131–146. 10.1016/j.cognition.2006.03.001 [DOI] [PubMed] [Google Scholar]
  12. Breedin SD, Saffran EM, & Coslett HB (1994). Reversal of the concreteness effect in a patient with semantic dementia. Cognitive Neuropsychology, 11(6), 617–660. [Google Scholar]
  13. Brysbaert M, & New B (2009). Moving beyond Kucera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41(4), 977–990. https://doi.org/41/4/977 [pii] 10.3758/BRM.41.4.977 [doi] [DOI] [PubMed] [Google Scholar]
  14. Brysbaert M, Stevens M, Mandera P, & Keuleers E (2016). How many words do we know? Practical estimates of vocabulary size dependent on word definition, the degree of language input and the participant’s age. Frontiers in Psychology, 29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Brysbaert M, Warriner AB, & Kuperman V (2014). Concreteness ratings for 40 thousand generally known English word lemmas. Behavior Research Methods, 46(3), 904–911. [DOI] [PubMed] [Google Scholar]
  16. Brysbaert M, Wijnendaele IV, & Deyne SD (2000). Age-of-acquisition effects in semantic processing tasks. Acta Psychologica, 104(2), 215–226. [DOI] [PubMed] [Google Scholar]
  17. Cabeza R. (2002). Hemispheric asymmetry reduction in older adults: The HAROLD model. Psychol Aging, 17(1), 85–100. [DOI] [PubMed] [Google Scholar]
  18. Cabeza R, & Dennis NA (2012). Frontal lobes and aging: Deterioration and compensation. Principles of Frontal Lobe Function, 2, 628–652. [Google Scholar]
  19. Cheng X, Popal H, Wang H, Hu R, Zang Y, Zhang M, Thornton MA, Cai H, Bi Y, Reilly J, Olson IR, & Wang Y (2023). The Conceptual Structure of Human Relationships Across Modern and Historical Cultures. OSF. 10.31234/osf.io/ut6qp [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Cohen G, & Faulkner D (1986). Memory for proper names: Age differences in retrieval. British Journal of Developmental Psychology, 4(2), 187–197. 10.1111/j.2044-835X.1986.tb01010.x [DOI] [Google Scholar]
  21. Coltheart M. (2000). Deep dyslexia is right-hemisphere reading. Brain and Language, 71(2), 299–309. [DOI] [PubMed] [Google Scholar]
  22. Coltheart M. (2006). Acquired dyslexias and the computational modelling of reading. Cognitive Neuropsychology, 23(1), 96–109. [DOI] [PubMed] [Google Scholar]
  23. Coltheart V, Laxon VJ, & Keating C (1988). Effects of word imageability and age of acquisition on children’s reading. British Journal of Psychology, 79(1), 1–12. [Google Scholar]
  24. Crutch SJ, & Warrington EK (2005). Abstract and concrete concepts have structurally different representational frameworks. Brain, 128(3), 615–627. [DOI] [PubMed] [Google Scholar]
  25. Cutler A. (1981). Making up materials is a confounded nuisance: Or Will we be able to run any psycholinguistic experiments at all in 1990? Cognition, 10(1–3), 65–70. [DOI] [PubMed] [Google Scholar]
  26. Dell’Acqua R, Sessa P, Peressotti F, Mulatti C, Navarrete E, & Grainger J (2010). ERP evidence for ultra-fast semantic processing in the picture–word interference paradigm. Frontiers in Psychology, 1, 177. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Dennis PA, & Hess TM (2016). Aging-related gains and losses associated with word production in connected speech. Aging, Neuropsychology, and Cognition, 23(6), 638–650. 10.1080/13825585.2016.1158233 [DOI] [PubMed] [Google Scholar]
  28. Dhooge E, & Hartsuiker RJ (2011). How do speakers resist distraction? Evidence from a taboo picture-word interference task. Psychological Science, 22(7), 855–859. 10.1177/0956797611410984 [DOI] [PubMed] [Google Scholar]
  29. Erickson KI, Miller DL, & Roecklein KA (2012). The Aging Hippocampus: Interactions between Exercise, Depression, and BDNF. The Neuroscientist, 18(1), 82–97. 10.1177/1073858410397054 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Fedorenko E. (2014). The role of domain-general cognitive control in language comprehension. Frontiers in Psychology, 5, 335. 10.3389/fpsyg.2014.00335 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Fedorenko E, Duncan J, & Kanwisher N (2012). Language-selective and domain-general regions lie side by side within Broca’s area. Current Biology, 22(21), 2059–2062. 10.1016/j.cub.2012.09.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Fedorenko E, Duncan J, & Kanwisher N (2013). Broad domain generality in focal regions of frontal and parietal cortex. Proceedings of the National Academy of Sciences, 110(41), 16616–16621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Fernandino L, Tong J-Q, Conant LL, Humphries CJ, & Binder JR (2022). Decoding the information structure underlying the neural representation of concepts. Proceedings of the National Academy of Sciences of the United States of America, 119(6), e2108091119. 10.1073/pnas.2108091119 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Festini SB, Zahodne L, & Reuter-Lorenz PA (2018). Theoretical Perspectives on Age Differences in Brain Activation: HAROLD, PASA, CRUNCH—How Do They STAC Up? In Oxford Research Encyclopedia of Psychology. https://oxfordre.com/psychology/psychology/psychology/abstract/10.1093/acrefore/9780190236557.001.0001/acrefore-9780190236557-e-400 [Google Scholar]
  35. Fini C, Era V, Da Rold F, Candidi M, & Borghi AM (2021). Abstract concepts in interaction: The need of others when guessing abstract concepts smooths dyadic motor interactions. Royal Society Open Science, 8(7), 201205. 10.1098/rsos.201205 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Fini C, Falcinelli I, Cuomo G, Era V, Candidi M, Tummolini L, Mazzuca C, & Borghi AM (2023). Breaking the ice in a conversation: Abstract words prompt dialogs more easily than concrete ones. Language and Cognition, 15(4), 629–650. 10.1017/langcog.2023.3 [DOI] [Google Scholar]
  37. Fjell AM, Sneve MH, Grydeland H, Storsve AB, & Walhovd KB (2017). The Disconnected Brain and Executive Function Decline in Aging. Cerebral Cortex, 27(3), 2303–2317. 10.1093/cercor/bhw082 [DOI] [PubMed] [Google Scholar]
  38. Franklin S, Howard D, & Patterson K (1994). Abstract word meaning deafness. Cognitive Neuropsychology, 11(1), 1–34. [Google Scholar]
  39. Franklin S, Howard D, & Patterson K (1995). Abstract word anomia. Cognitive Neuropsychology, 12(5), 549–566. [Google Scholar]
  40. Gagolewski M, & Tartanus B (2023). stringi: Fast and Portable Character String Processing Facilities (Version 1.7.12) [Computer software]. https://cran.r-project.org/web/packages/stringi/index.html [Google Scholar]
  41. Gandolfi G, Pickering MJ, & Garrod S (2023). Mechanisms of alignment: Shared control, social cognition and metacognition. Philosophical Transactions of the Royal Society B: Biological Sciences, 378(1870), 20210362. 10.1098/rstb.2021.0362 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Garrard P, Lambon Ralph M. a, Patterson K, Pratt KH, & Hodges JR (2005). Semantic feature knowledge and picture naming in dementia of Alzheimer’s type: A new approach. Brain and Language, 93(1), 79–94. https://doi.org/S0093-934X(04)00226-3 [pii] 10.1016/j.bandl.2004.08.003 [doi] [DOI] [PubMed] [Google Scholar]
  43. Garrard P, Maloney LM, Hodges JR, & Patterson K (2005). The effects of very early Alzheimer’s disease on the characteristics of writing by a renowned author. Brain, 128(Pt 2), 250–260. https://doi.org/awh341 [pii] 10.1093/brain/awh341 [doi] [DOI] [PubMed] [Google Scholar]
  44. Gleitman LR, Cassidy K, Nappa R, Papafragou A, & Trueswell JC (2021). Hard words. In Language in development: A crosslinguistic perspective (pp. 33–77). The MIT Press. 10.7551/mitpress/11865.003.0006 [DOI] [Google Scholar]
  45. Goral M, Spiro Iii A, Albert ML, Obler LK, & Connor LT (2007). Change in lexical retrieval skills in adulthood. The Mental Lexicon, 2(2), 215–238. 10.1075/ml.2.2.05gor [DOI] [Google Scholar]
  46. Grice HP (1975). Logic and conversation. In Cole P & Morgan JL (Eds.), Syntax and Semantics 3. Speech Acts (pp. 41–58). Academic Press. [Google Scholar]
  47. Hartshorne JK, & Germine LT (2015). When Does Cognitive Functioning Peak? The Asynchronous Rise and Fall of Different Cognitive Abilities Across the Life Span. Psychological Science, 26(4), 433–443. 10.1177/0956797614567339 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Heller D, & Brown-Schmidt S (2023). The Multiple Perspectives Theory of Mental States in Communication. Cognitive Science, 47(7), e13322. 10.1111/cogs.13322 [DOI] [PubMed] [Google Scholar]
  49. Hernandez AE, Ronderos J, Bodet JP, Claussenius-Kalman H, Nguyen MV, & Bunta F (2021). German in childhood and Latin in adolescence: On the bidialectal nature of lexical access in English. Humanities and Social Sciences Communications, 8(1), 1–12. 10.1057/s41599-021-00836-438617731 [DOI] [Google Scholar]
  50. Higby E, Cahana-Amitay D, Vogel-Eyny A, Spiro III A, Albert ML, & Obler LK (2019). The Role of Executive Functions in Object- and Action-Naming among Older Adults. Experimental Aging Research, 45(4), 306–330. 10.1080/0361073X.2019.1627492 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Hodges JR, Graham N, & Patterson K (1995). Charting the progression in semantic dementia: Implications for the organisation of semantic memory. Memory, 3, 463–495. [DOI] [PubMed] [Google Scholar]
  52. Hoffman P, Jefferies E, & Lambon Ralph MA (2011). Remembering “zeal” but not “thing”: Reverse frequency effects as a consequence of deregulated semantic processing. Neuropsychologia, 49(3), 580–584. https://doi.org/S0028-3932(10)00580-4 [pii] 10.1016/j.neuropsychologia.2010.12.036 [doi] [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Hoffman P, Jefferies E, & Ralph MAL (2010). Ventrolateral prefrontal cortex plays an executive regulation role in comprehension of abstract words: Convergent neuropsychological and repetitive TMS evidence. Journal of Neuroscience, 30(46), 15450–15456. 10.1523/JNEUROSCI.3783-10.2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Hoffman P, Jones RW, & Lambon Ralph MA (2013). Be concrete to be comprehended: Consistent imageability effects in semantic dementia for nouns, verbs, synonyms and associates. Cortex, 49(5), 1206–1218. [DOI] [PubMed] [Google Scholar]
  55. Hoffman P, & Lambon Ralph MA (2011). Reverse concreteness effects are not a typical feature of semantic dementia: Evidence for the hub-and-spoke model of conceptual representation. Cerebral Cortex, bhq288. [DOI] [PubMed] [Google Scholar]
  56. Hoffman P, Lambon Ralph MA, & Rogers TT (2013). Semantic diversity: A measure of semantic ambiguity based on variability in the contextual usage of words. Behavior Research Methods, 45(3), 718–730. https://doi.org/10/f5d59g [DOI] [PubMed] [Google Scholar]
  57. Hoffman P, McClelland JL, & Lambon Ralph MA (2018). Concepts, control, and context: A connectionist account of normal and disordered semantic cognition. Psychological Review, 125(3), 293–328. 10.1037/rev0000094 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Hoffman P, Meteyard L, & Patterson K (2014). Broadly speaking: Vocabulary in semantic dementia shifts towards general, semantically diverse words. Cortex, 55, 30–42. 10.1016/j.cortex.2012.11.004 [DOI] [PubMed] [Google Scholar]
  59. Hoffman P, & Morcom AM (2018). Age-related changes in the neural networks supporting semantic cognition: A meta-analysis of 47 functional neuroimaging studies. Neuroscience & Biobehavioral Reviews, 84, 134–150. 10.1016/j.neubiorev.2017.11.010 [DOI] [PubMed] [Google Scholar]
  60. Horton WS (2005). Conversational common ground and memory processes in language production. Discourse Processes, 40(1), 1–35. 10.1207/s15326950dp4001_1 [DOI] [Google Scholar]
  61. Horton WS, Spieler DH, & Shriberg E (2010). A corpus analysis of patterns of age-related change in conversational speech. Psychology and Aging, 25(3), 708–713. 10.1037/a0019424 [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Hothorn T, Zeileis A, Farebrother (pan.f) RW, Cummins (pan.f) C, Millo G, & Mitchell D (2022). lmtest: Testing Linear Regression Models (Version 0.9-40) [Computer software]. https://cran.r-project.org/web/packages/lmtest/index.html [Google Scholar]
  63. Jackson RL (2021). The neural correlates of semantic control revisited. NeuroImage, 224, 117444. 10.1016/j.neuroimage.2020.117444 [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Jackson RL, Rogers TT, & Lambon Ralph MA (2021). Reverse-engineering the cortical architecture for controlled semantic cognition. Nature Human Behaviour, 5(6), 774–786. 10.1038/s41562-020-01034-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Jefferies E, Patterson K, & Lambon Ralph MA (2008). Deficits of knowledge versus executive control in semantic cognition: Insights from cued naming. 46, 649–658. 10.1016/j.neuropsychologia.2007.09.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Juhasz BJ (2005). Age-of-acquisition effects in word and picture identification. Psychological Bulletin, 131(5), 684–712. [DOI] [PubMed] [Google Scholar]
  67. Kavé G. (2024). Vocabulary changes in adulthood: Main findings and methodological considerations. International Journal of Language & Communication Disorders, 59(1), 58–67. 10.1111/1460-6984.12820 [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Kavé G, & Goral M (2017). Do age-related word retrieval difficulties appear (or disappear) in connected speech? Neuropsychology, Development, and Cognition. Section B, Aging, Neuropsychology and Cognition, 24(5), 508–527. 10.1080/13825585.2016.1226249 [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Kearney E, McMahon KL, Guenther F, Arciuli J, & de Zubicaray GI (2025). Revisiting the concreteness effect: Non-arbitrary mappings between form and concreteness of English words influence lexical processing. Cognition, 254, 105972. 10.1016/j.cognition.2024.105972 [DOI] [PubMed] [Google Scholar]
  70. Kemper S, Marquis J, & Thompson M (2001). Longitudinal change in language production: Effects of aging and dementia on grammatical complexity and propositional content. Psychology and Aging, 16(4), 600–614. [DOI] [PubMed] [Google Scholar]
  71. Kita S, & Özyürek A (2003). What does cross-linguistic variation in semantic coordination of speech and gesture reveal?: Evidence for an interface representation of spatial thinking and speaking. Journal of Memory and Language, 48(1), 16–32. 10.1016/S0749-596X(02)00505-3 [DOI] [Google Scholar]
  72. Kousta S-TT, Vigliocco G, Vinson DP, Andrews M, & Del Campo E (2011). The representation of abstract words: Why emotion matters. Journal of Experimental Psychology General, 140(1), 14–34. https://doi.org/2010-26153-001 [pii] 10.1037/a0021446 [doi] [DOI] [PubMed] [Google Scholar]
  73. Lievers FS, Bolognesi M, & Winter B (2021). The linguistic dimensions of concrete and abstract concepts: Lexical category, morphological structure, countability, and etymology. Cognitive Linguistics, 32(4), 641–670. 10.1515/cog-2021-0007 [DOI] [Google Scholar]
  74. Liu B (2020). Text sentiment analysis based on CBOW model and deep learning in big data environment. Journal of Ambient Intelligence and Humanized Computing, 11(2), 451–458. 10.1007/s12652-018-1095-6 [DOI] [Google Scholar]
  75. Lüdecke D, Bartel A, Schwemmer C, Powell C, Djalovski A, & Titz J (2024). sjPlot: Data Visualization for Statistics in Social Science (Version 2.8.17) [Computer software]. https://cran.r-project.org/web/packages/sjPlot/index.html [Google Scholar]
  76. Lupker SJ (1979). The semantic nature of response competition in the picture-word interference task. Memory & Cognition, 7(6), 485–495. [Google Scholar]
  77. Macoir J. (2009). Is a plum a memory problem?: Longitudinal study of the reversal of concreteness effect in a patient with semantic dementia. Neuropsychologia, 47(2), 518–535. [DOI] [PubMed] [Google Scholar]
  78. Maylor EA (1990). Recognizing and naming faces: Aging, memory retrieval, and the tip of the tongue state. Journal of Gerontology, 45(6), P215–P226. [DOI] [PubMed] [Google Scholar]
  79. Mazzuca C, Villani C, Lamarra T, Bolognesi MM, & Borghi AM (2025). Abstractness impacts conversational dynamics. Cognition, 258, 106084. 10.1016/j.cognition.2025.106084 [DOI] [PubMed] [Google Scholar]
  80. Menenti L, Pickering MJ, & Garrod SC (2012). Toward a neural basis of interactive alignment in conversation. Frontiers in Human Neuroscience, 6, 185. 10.3389/fnhum.2012.00185 [DOI] [PMC free article] [PubMed] [Google Scholar]
  81. Mohammad S. (2018). Obtaining Reliable Human Ratings of Valence, Arousal, and Dominance for 20,000 English Words. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 174–184. 10.18653/v1/P18-1017 [DOI] [Google Scholar]
  82. Moscovitch M, & Winocur G (1995). Frontal lobes, memory, and aging. Annals-New York Academy of Sciences, 769, 119–150. 10.1111/j.1749-6632.1995.tb38135.x [DOI] [PubMed] [Google Scholar]
  83. Noppeney U, & Price CJ (2004). Retrieval of abstract semantics. NeuroImage, 22(1), 164–170. 10.1016/j.neuroimage.2003.12.010 [DOI] [PubMed] [Google Scholar]
  84. Paivio A. (1991). Dual coding theory: Retrospect and current status. Canadian Journal of Psychology, 45(3), 255–287. [Google Scholar]
  85. Paivio A, Walsh M, & Bons T (1994). Concreteness effects on memory: When and why? Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(5), 1196–1204. [Google Scholar]
  86. Papagno C, Fogliata A, Catricala E, & Miniussi C (2009). The lexical processing of abstract and concrete nouns. Brain Research, 1263, 78–86. https://doi.org/S0006-8993(09)00182-6 [pii] 10.1016/j.brainres.2009.01.037 [doi] [DOI] [PubMed] [Google Scholar]
  87. Park DC, & Reuter-Lorenz P (2009). The adaptive brain: Aging and neurocognitive scaffolding. Annual Review of Psychology, 60, 173–196. 10.1146/annurev.psych.59.103006.093656 [DOI] [PMC free article] [PubMed] [Google Scholar]
  88. Pexman PM, Heard A, Lloyd E, & Yap MJ (2016). The Calgary semantic decision project: Concrete/abstract decision data for 10,000 English words. Behavior Research Methods, 1–11. 10.3758/s13428-016-0720-6 [DOI] [PubMed] [Google Scholar]
  89. Pickering MJ, & Garrod S (2004). Toward a mechanistic psychology of dialogue. The Behavioral and Brain Sciences, 27(2), 169–190; discussion 190-226. [DOI] [PubMed] [Google Scholar]
  90. Pinker S. (1995). Language acquisition. Gleitman Lila R., 1(2nd ed), 135–182. [Google Scholar]
  91. Raji S, & da Melo G (2020). What sparks joy: The AffectVec emotion database. Proceedings of the Web Conference, ACM. [Google Scholar]
  92. Ramey CH, Chrysikou EG, & Reilly J (2013). Snapshots of Children’s Changing Biases During Language Development: Differential Weighting of Perceptual and Linguistic Factors Predicts Noun Age of Acquisition. Journal of Cognition and Development, 14(4), 573–592. 10.1080/15248372.2012.689386 [DOI] [Google Scholar]
  93. Rasenberg M, Özyürek A, & Dingemanse M (2020). Alignment in multimodal interaction: An integrative framework. Cognitive Science, 44(11), e12911. 10.1111/cogs.12911 [DOI] [PMC free article] [PubMed] [Google Scholar]
  94. Reece A, Cooney G, Bull P, Chung C, Dawson B, Fitzpatrick C, Glazer T, Knox D, Liebscher A, & Marin S (2023). The CANDOR corpus: Insights from a large multimodal dataset of naturalistic conversation. Science Advances, 9(13), eadf3197. 10.1126/sciadv.adf3197 [DOI] [PMC free article] [PubMed] [Google Scholar]
  95. Reilly J, Cross K, Troiani V, & Grossman M (2007). Single-word semantic judgements in semantic dementia: Do phonology and grammatical class count? Aphasiology, 21(6–8), 558–569. 10.1080/02687030701191986 [DOI] [Google Scholar]
  96. Reilly J, Hung J, & Westbury C (2017). Non-Arbitrariness in mapping word form to meaning: Cross-linguistic formal markers of word concreteness. Cognitive Science, 41(4), 1071–1089. [DOI] [PubMed] [Google Scholar]
  97. Reilly J, & Kean J (2007). Formal distinctiveness of high- and low-imageability nouns: Analyses and theoretical implications. Cognitive Science, 31(1), 157–168. 10.1080/03640210709336988 [DOI] [PubMed] [Google Scholar]
  98. Reilly J, Peelle JE, Antonucci SM, & Grossman M (2011). Anomia as a marker of distinct semantic memory impairments in Alzheimer’s disease and semantic dementia. Neuropsychology, 25(4), 413–426. 10.1037/a0022738 [DOI] [PMC free article] [PubMed] [Google Scholar]
  99. Reilly J, Peelle JE, Garcia A, & Crutch SJ (2016). Linking somatic and symbolic representation in semantic memory: The dynamic multilevel reactivation framework. Psychonomic Bulletin & Review, 23(4), 1002–1014. 10.3758/s13423-015-0824-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  100. Reilly J, Shain C, Borghesani V, Kuhnke P, Vigliocco G, Peelle JE, Mahon BZ, Buxbaum LJ, Majid A, Brysbaert M, Borghi AM, De Deyne S, Dove G, Papeo L, Pexman PM, Poeppel D, Lupyan G, Boggio P, Hickok G, … Vinson D (2024). What we mean when we say semantic: Toward a multidisciplinary semantic glossary. Psychonomic Bulletin & Review. 10.3758/s13423-024-02556-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  101. Reilly J, Shain C, Borghesani V, Kuhnke P, Vigliocco G, Peelle JE, Mahon BZ, Buxbaum LJ, Majid A, Brysbaert M, Borghi AM, De Deyne S, Dove G, Papeo L, Pexman PM, Poeppel D, Lupyan G, Boggio P, Hickok G, … Vinson D (2025). What we mean when we say semantic: Toward a multidisciplinary semantic glossary. Psychonomic Bulletin & Review, 32(1), 243–280. 10.3758/s13423-024-02556-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  102. Reilly J, Ulichney V, Sacks B, Duncan A, Weinstein T, Giovannetti T, Helion C, & Cooney G (2023). ConversationAlign: A Computational pipeline for evaluating alignment in natural language dyadic interactions. OSF. 10.31234/osf.io/3crua [DOI] [Google Scholar]
  103. Reuter-Lorenz PA, & Cappell KA (2008). Neurocognitive Aging and the Compensation Hypothesis. Current Directions in Psychological Science, 17(3), 177–182. 10.1111/j.1467-8721.2008.00570.x [DOI] [Google Scholar]
  104. Rinker TW (2018a). Textstem: Tools for stemming and lammatizing (Version 0.1.4) [Computer software]. [Google Scholar]
  105. Rinker TW (2018b). Textclean (Version 0.9.3) [Computer software]. https://CRAN.R-project.org/package=textclean [Google Scholar]
  106. Rönnlund M, Nyberg L, Bäckman L, & Nilsson L-G (2005). Stability, growth, and decline in adult life span development of declarative memory: Cross-sectional and longitudinal data from a population-based study. Psychology and Aging, 20(1), 3. 10.1037/0882-7974.20.1.3 [DOI] [PubMed] [Google Scholar]
  107. Sacks B, Ulichney V, Duncan A, Helion C, Weinstein S, Giovannetti T, Cooney G, & Reilly J (2025). ConversationAlign: Open-Source Software for Analyzing Patterns of Lexical Use andAlignment in Conversation Transcripts. PsyArXiv. https://osf.io/preprints/psyarxiv/7xqnp_v1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  108. Salthouse TA (2009). When does age-related cognitive decline begin? Neurobiology of Aging, 30(4), 507–514. 10.1016/j.neurobiolaging.2008.09.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  109. Sandberg C, & Kiran S (2014). How justice can affect jury: Training abstract words promotes generalisation to concrete words in patients with aphasia. Neuropsychological Rehabilitation, 24(5), 738–769. [DOI] [PMC free article] [PubMed] [Google Scholar]
  110. Schriefers H, Meyer AS, & Levelt WJ (1990). Exploring the time course of lexical access in language production: Picture-word interference studies. Journal of Memory and Language, 29(1), 86–102. 10.1016/0749-596X(90)90011-N [DOI] [Google Scholar]
  111. Shafto MA, & Tyler LK (2014). Language in the aging brain: The network dynamics of cognitive decline and preservation. Science, 346(6209), 583–587. 10.1126/science.1254404 [DOI] [PubMed] [Google Scholar]
  112. Sikora E, Bielak-Zmijewska A, Dudkowska M, Krzystyniak A, Mosieniak G, Wesierska M, & Wlodarczyk J (2021). Cellular Senescence in Brain Aging. Frontiers in Aging Neuroscience, 13. 10.3389/fnagi.2021.646924 [DOI] [PMC free article] [PubMed] [Google Scholar]
  113. Stalnaker R (2002). Common ground. Linguistics and Philosophy, 25(5/6), 701–721. 10.1023/A:1020867916902 [DOI] [Google Scholar]
  114. Tapia-Arancibia L, Aliaga E, Silhol M, & Arancibia S (2008). New insights into brain BDNF function in normal aging and Alzheimer disease. Brain Research Reviews, 59(1), 201–220. 10.1016/j.brainresrev.2008.07.007 [DOI] [PubMed] [Google Scholar]
  115. Tomoeda CK, Bayles KA, Trosset MW, Azuma T, & al E (1996). Cross-sectional analysis of Alzheimer disease effects on oral discourse in a picture description task. Alzheimer Disease & Associated Disorders, 10(4), 204–215. [DOI] [PubMed] [Google Scholar]
  116. Troche J, Crutch S, & Reilly J (2014). Clustering, hierarchical organization, and the topography of abstract and concrete nouns. Frontiers in Psychology, 5, 360. 10.3389/fpsyg.2014.00360 [DOI] [PMC free article] [PubMed] [Google Scholar]
  117. van Paridon JP, Bolker BM, & Alday P (2023). lmerMultiMember: Multiple membership random effects. R Package Version 0.11, 8. [Google Scholar]
  118. Vigliocco G, Kousta S-T, Della Rosa PA, Vinson DP, Tettamanti M, Devlin JT, & Cappa SF (2014). The neural representation of abstract words: The role of emotion. Cerebral Cortex (New York, N.Y. : 1991), 24(7), 1767–1777. 10.1093/cercor/bht025 [DOI] [PubMed] [Google Scholar]
  119. Villani C, Lugli L, Liuzza MT, Nicoletti R, & Borghi AM (2021). Sensorimotor and interoceptive dimensions in concrete and abstract concepts. Journal of Memory and Language, 116, 104173. 10.1016/j.jml.2020.104173 [DOI] [PMC free article] [PubMed] [Google Scholar]
  120. Villani C, Orsoni M, Lugli L, Benassi M, & Borghi AM (2022). Abstract and concrete concepts in conversation. Scientific Reports, 12(1), 17572. 10.1038/s41598-022-20785-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  121. Vinke EJ, De Groot M, Venkatraghavan V, Klein S, Niessen WJ, Ikram MA, & Vernooij MW (2018). Trajectories of imaging markers in brain aging: The Rotterdam Study. Neurobiology of Aging, 71, 32–40. 10.1016/j.neurobiolaging.2018.07.001 [DOI] [PubMed] [Google Scholar]
  122. Walker I, & Hulme C (1999). Concrete words are easier to recall than abstract words: Evidence for a semantic contribution to short-term serial recall. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25(5), 1256–1271. [Google Scholar]
  123. Wierenga CE, Benjamin M, Gopinath K, Perlstein WM, Leonard CM, Rothi LJ, Conway T, Cato MA, Briggs R, & Crosson B (2008). Age-related changes in word retrieval: Role of bilateral frontal and subcortical networks. Neurobiology of Aging, 29(3), 436–451. https://doi.org/S0197-4580(06)00395-2 [pii] 10.1016/j.neurobiolaging.2006.10.024 [doi] [DOI] [PubMed] [Google Scholar]
  124. Wijffels J. (2023). udpipe: Tokenization, Parts of Speech Tagging, Lemmatization and Dependency Parsing with the “UDPipe” “NLP” Toolkit (Version R package version 0.8.11) [Computer software]. [Google Scholar]
  125. Williams A, & Nussbaum JF (2013). Intergenerational communication across the life span. Routledge. [Google Scholar]
  126. Willroth EC, & Atherton OE (2024). Best laid plans: A Guide to reporting preregistration deviations. Advances in Methods and Practices in Psychological Science, 7(1), 25152459231213802. 10.1177/25152459231213802 [DOI] [Google Scholar]
  127. Woollams AM, Cooper-Pye E, Hodges JR, & Patterson K (2008). Anomia: A doubly typical signature of semantic dementia. Neuropsychologia, 46(10), 2503–2514. [DOI] [PubMed] [Google Scholar]
  128. Xu Y, & Reitter D (2015). An evaluation and comparison of linguistic alignment measures. Proceedings of the 6th Workshop on Cognitive Modeling and Computational Linguistics, 58–67. 10.3115/v1/W15-1107 [DOI] [Google Scholar]
  129. Yap MJ, & Balota DA (2009). Visual word recognition of multisyllabic words. Journal of Memory and Language, 60(4), 502–529. [Google Scholar]
  130. Yi H, Moore P, & Grossman M (2007). Reversal of the concreteness effect for verbs in semantic dementia. Cognitive Neuropsychology, 21(1), 1–19. [DOI] [PubMed] [Google Scholar]
  131. Zdrazilova L, Sidhu DM, & Pexman PM (2018). Communicating abstract meaning: Concepts revealed in words and gestures. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1752), 20170138. 10.1098/rstb.2017.0138 [DOI] [PMC free article] [PubMed] [Google Scholar]
  132. Zhang H, & Diaz MT (2023). Task difficulty modulates age-related differences in functional connectivity during word production. Brain and Language, 240, 105263. 10.1016/j.bandl.2023.105263 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

1
NIHMS2099584-supplement-1.docx (130.1KB, docx)

Data Availability Statement

All data and associated code are available via the Open Science Foundation at https://osf.io/atf5q/?view_only=3abcf674d398421eb2d4f9d3e88e6063

RESOURCES