Word learning in early, middle and late adulthood

Lead Research Organisation: University of York
Department Name: Psychology


In the field of cognitive aging, it is well established that some cognitive functions remain relatively stable over the lifespan (e.g., long-term representations, automatic processes, implicit learning) while others progressively decline (e.g., executive control, memory maintenance/retrieval, and speed). Language use shows a mixed aging trajectory as well: Phonological, lexical, semantic and syntactic knowledge is relatively preserved, whereas the ease of access to that knowledge and control over its activation tend to degrade with age. How this dissociation affects the dynamics and efficiency of word learning over the lifespan is unclear.

The purpose of this research programme is to investigate the age-related stability vs. decline of several key word-learning mechanisms that have been documented in the word-acquisition literature in infancy and childhood, and to establish the extent to which compensatory trade-offs (e.g., reliance on stored knowledge) can stave off the decline of learning mechanisms that involve a greater degree of executive control and explicit computation. A better understanding of how the mechanisms underpinning word learning change over the lifespan has important consequences for older adults' quality of life. Indeed, owing to today's increased life expectancy and the expectation that many of those additional years will be spent in relatively good health, opportunities for older adults to acquire new knowledge, new skills, or learn new languages have flourished.

The target groups will be young adults (< 25y), middle-age adults (40y to 50y), and older adults (> 65y). The middle-age group is critical because it will enable us not only to draw a more accurate trajectory of word learning mechanisms, but also to establish whether compensation based on context and long-term representations begins to operate earlier than traditionally assumed. The learning mechanisms under study will be: (1) Adjacent dependencies: This represents a form of domain-general learning based on implicitly computing statistical regularities in the speech input. We expect this type of learning to be comparatively robust over time. (2) Non-adjacent dependencies: This will instantiate a form of word learning that relies on both association and abstraction, a more complex form of learning that could potentially engage processing resources that are in short supply in older adults. (3) Abstract patterns: These experiments will instantiate abstract pattern learning and generalisation to entirely new materials, unlike adjacent dependencies which require no generalisation to new materials and non-adjacent dependencies which require only partial generalisation to new materials. (4) Rule-based abstraction (first-order logic): This will be similar to the previous set of experiments, except that the abstract patterns defining the acceptable words of the language will be regulated by a grammar containing logical operators (equivalent to if, then, else). This is the most language-specific form of learning of the four sets. Its resilience to aging is largely unknown.

The above learning mechanisms will be compared to their matched explicit-learning counterpart. As a whole, this study will highlight the dynamics and plasticity of language learning over the lifespan and, in doing so, will contribute to our theoretical understanding of language processing, language learning, and cognitive aging.

Planned Impact

The research is fundamental science and its primary impact will be scientific. However, the advances in theory and methodology will benefit researchers and practitioners in non-academic settings such as clinical practice, educators, and others with an interest in improving quality of life in older adults. The postdoctoral research fellow's and research assistant's careers will also benefit from the project.

Owing to today's increased life expectancy and the expectation that many of those additional years will be spent in relatively good health, opportunities for older adults to acquire new knowledge, new skills, or learn new languages have flourished (e.g., The University of the Third Age, www.u3a.org.uk). Therefore, our research has the potential to contribute to widening participation in higher education through groups like, e.g., the University of York Centre for Lifelong Learning (http://www.york.ac.uk/lifelonglearning/).

Maintaining the skills that support language learning in old age is crucial not only for second-language learning, but also for reducing older adults' communicative isolation and a sense of losing touch with their own native language, especially with words and expressions pertaining to new technologies such as the Internet and wireless communication. Thus, our research will address fundamental questions such as: How can we keep word learning effective in old age? Can age-related language deficiencies be managed or even attenuated through compensatory mechanisms?

Answers to these questions may facilitate clinical assessment of age-related language declines associated with fronto-temporal dementia and Alzheimer's disease, and especially help identify language-learning skills that might have been spared by the disease (e.g., implicit word learning). Establishing the extent to which reliance on lexical context and long-term memory can compensate for learning difficulties could also inform intervention practice and rehabilitation.

Our work should also be of interest to second-language teachers focusing on teaching foreign languages to mature students. Our data will provide valuable evidence on how to improve older learners' experience of language learning, especially when sensory decline is observed. In that context, the PI's active involvement in the FP7 Marie Curie Initial Training Network INSPIRE (INvestigating Speech Processing in Realistic Environments) will serve as a platform for liaising with the industrial and clinical partners of the network on how to best integrate language learning in the elderly with speech perception and hearing intervention (e.g., Otikon, Phonak, Cochlear, Royal National Throat, Nose and Ear Hospital, Radboud Medical Hospital Center).

On a broader level, we expect that this research will be of relevance to policy makers and the general public. The steady increase in the pensionable age means that adults in the UK are now expected to remain active in the workforce for longer than previous generations. This coincides with a rapid change in the labour market, which demands greater flexibility from employees, and in which a "job for life" is no longer a reality. The implication is that there is increasing pressure for middle-aged adults to re-train and learn new skills. Thus, understanding how the mechanisms that underpin learning and memory change throughout life is essential if we are to support efficient and effective learning in an aging population.


10 25 50
publication icon
Palmer SD (2019) Lexical knowledge boosts statistically-driven speech segmentation. in Journal of experimental psychology. Learning, memory, and cognition

publication icon
Palmer SD (2018) Lexical knowledge boosts statistically-driven speech segmentation in Journal of Experimental Psychology: Learning, Memory, and Cognition

publication icon
Palmer SD (2016) Speech segmentation by statistical learning is supported by domain-general processes within working memory. in Quarterly journal of experimental psychology (2006)

Description The project aimed to: (1) Investigate the trajectory across adulthood of well-documented language-learning mechanisms, with a focus on statistical learning and non-adjacent dependencies, (2) Identify the mechanisms that are more vs. less resilient to aging, (3) Evaluate if resilience to aging is related to the extent to which those mechanisms involve executive control vs. stored knowledge, and (4) Test whether learning mechanisms that are particularly sensitive to age can be compensated for by long-term knowledge.

Key results:
(1) Statistical learning shows broad consistency across the lifespan
We exposed young, middle-aged, and older adults (N = 32 in each age group) to artificial languages (uninterrupted speech streams) with and without cognitive load. The cognitive load consisted of a visual 2-back task to be performed while listening to the stream. Language learning was assessed immediately after stream exposure. Participants also completed a battery of cognitive tests assessing memory, attention, and reasoning. The results showed that statistical learning was remarkably resilient to ageing, although the older adults showed poorer performance under cognitive load. This result points to age-related stability in implicit learning, but limitations in allocating cognitive resources in the older group.

(2) Statistical learning relies on a combination of updating processes and working memory capacity
Modelling the above data sets based on individual psychometric scores showed that statistical learning is predicted by memory updating, whereas a simpler, familiarity-based measure of learning is predicted by working memory storage capacity. These two capacities were also found to differ across age groups. In particular, older individuals had more difficulty updating information in working memory compared to younger individuals.

(3) Lexical knowledge boosts statistically-driven speech segmentation
We investigated the consequences on statistical learning of introducing a known word in the artificial language. With young adults, we found that the presence of a known word (e.g., tomorrow) improved segmentation of the novel words. This improvement was maintained when the real word was a different length form the novel words (philosophy), ruling out an explanation based on metrical expectations. These results show that recognizable portions of speech in an otherwise meaningless stream serve as anchors for discovering new words. We are currently testing older adults on a comparable set-up. Our initial hypothesis is that older adults might benefit more from known words because of their documented reliance on long-term knowledge (lexical information) during perceptual tasks. Surprisingly, the data so far indicate that older participants do not benefit from hearing real words. In fact, we found the opposite effect. We are currently investigating the possibility that older adults might struggle to disengage their attention from known words, leading to what is known as an "attentional blink," where the attentional capture caused by the known words disrupts encoding of the immediately following portions of speech, and hence, statistical learning.

(4) Learning non-adjacent dependencies across the lifespan
In this part of the work, we aimed to simulate a form of word learning that relies on both associative and abstract processes. During an exposure phase, participants heard quadrisyllabic phrases following an AXB format, where the middle disyllable (X) was free to vary, e.g., "pel-balip-vot," "pel-kicey-vot." At test, a novel instance of that pattern was tested, e.g., "pel-gooka-vot." This type of learning is thought to assess sensitivity to morphological derivation, and hence, it can be seen as a more complex linguistic skill than the statistical learning. Experiments so far have provided mixed results. Conditions in which the exposure phrases were separated by relatively brief intervals (300 or 400 ms) showed no learning in either age group. However, with longer intervals (500 and 750 ms), young participants showed learning. We are currently testing older participants. Our reasoning is that older participants might find it difficult to learn non-adjacent dependencies because extracting abstract patterns requires greater processing resource than extracting simple adjacent dependencies. However, we also hypothesise that increasing the intervals further (> 1000 ms) might allow older people to explicitly rehearse the phrases, and hence, show learning. The overall patterns of the non-adjacent and adjacent dependencies will be compared. This will allow us to identify the conditions under which age-related decline in word learning are best overcome.
Exploitation Route These experiments will be used to address the following questions: (1) Are young and older adults learning adjacent and non-adjacent dependencies differently in incidental vs. intentional conditions? (2) Is the lexical boost in young adults due to selective attention to (and learning of) the word immediately following the known word or does it generalise to all words of the language? (3) Is it possible to devise a new paradigm that can capture statistical learning without the limitations of conventional procedures (e.g., only provide crude accuracy data, heavily influenced by the nature of the test foils, are not conducive to within-subjects designs)?

Some of the results from this work can be found in: Palmer, S.L., Hutson, J., & Mattys, S.L. (2018). Statistical learning for speech segmentation: Age-related changes and underlying mechanisms. Psychology and Aging.
Sectors Communities and Social Services/Policy,Healthcare

URL https://pure.york.ac.uk/portal/en/publications/statistical-learning-for-speech-segmentation(6fe52507-b0c9-4ce7-ae1a-d62c7cb7b095).html
Description Research grants, new investigator (Ronan McGarrigle)
Amount £228,835 (GBP)
Funding ID ES/R003572/1 
Organisation Economic and Social Research Council 
Sector Public
Country United Kingdom
Start 10/2018 
End 02/2021
Description Collaboration with Mel Ferguson and NIHR 
Organisation National Institute for Health Research
Department NIHR Nottingham Biomedical Research Centre
Country United Kingdom 
Sector Academic/University 
PI Contribution Provided data on sound/speech perception in young and older listeners under cognitive load. This, in turn, was the impetus for subsequent collaborations in cognitive listening -- ESRC ES/4004722/1.
Collaborator Contribution Provided research and clinical expertise and advice for possible application and intervention. This, in turn, was the impetus for subsequent collaborations in cognitive listening -- ESRC ES/4004722/1.
Impact Newly funhded programme of research on cognitive listening, ESRC ES/4004722/1. Mostly psycholinguistic, but with ties to clinical audiology.
Start Year 2015
Description Invited address at the British Society of Audiologists meeting 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact Presentation to an audience of audiologists and hearing scientists of recent results on the effect of aging and cognitive load on speech perception and hearing processes.
Year(s) Of Engagement Activity 2016
URL http://www.thebsa.org.uk/overview-bsa-2016-annual-conference/
Description Open day for older participants 
Form Of Engagement Activity Participation in an open day or visit at my research institution
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Study participants or study members
Results and Impact Open day for older participants who volunteered in our research over the past two years. The event (Saturday morning) involved refreshments, networking, and a 90-minute presentation on psychology and aging. This was followed by a 30-min Q&A session.
Year(s) Of Engagement Activity 2018