This collection contains the quantitative data resulting from the analysis of the elderLUCID audio corpus – a set of speech recordings collected for 83 adults aged 19 to 84 years inclusive. Recordings were made while participants carried out two types of collaborative tasks with a conversational partner who was a young adult of the same sex: (1) a ‘spot the difference’ picture task (‘diapix’) where the conversational partners had to collaborate to find 12 differences between their pictures and (2) a BKB sentence repetition task where the key participant had to read a set of sentences to their partner who had to repeat them back. The two tasks were carried out by each participant pair in four different conditions: (1) in good listening conditions when both could hear each other normally (NORM condition), or when perception was impaired for one or both of the participants by (2) simulating a severe-to-profound hearing loss in the conversational partner (HLS condition), (3) adding multispeaker babble noise to the audio channel for the conversational partner (BAB1 condition) or to the audio channel of both participants (BAB2 condition). The aim of the study was to examine the clarification strategies used by older adults and young adult controls to maintain effective communication in adverse communicative conditions. The SPSS spreadsheet contains, for each of the 83 participants, quantitative data resulting from (a) the acoustic analysis of the recordings, (b) measures of communication efficiency and (c) background auditory and cognitive measures. Speech communication can be difficult for older people, due to the combined effects of age-related hearing loss, which is common over the age of 65, age-related decline in the quality of phonation and speech articulation, and cognitive problems such as poorer short-term memory and processing speed. Past studies of how older individuals perceive and produce speech sounds have tended to consider these abilities independently of each other using controlled materials, such as read words or sentences. These studies tell us little about how older speakers function when using speech for communicative purposes, and how these various factors interact. For example, it has been shown that older people benefit from seeing their interlocutor in conversations, but audiovisual speech places greater demands on cognitive processing than auditory speech which leads to increased listener effort and reduced information recall. In our project, we propose to gain a comprehensive account of older people's speech production and perception in situations involving communication with another individual. Adults with age-related hearing loss and the rarer group of older adults with normal hearing will be included as well as younger adult controls. In Study 1, communication with another speaker, while reading sentences or completing a problem-solving task, will either be in good listening conditions, where both speakers hear each other normally, or in adverse conditions, where the participant has to get their message across to another speaker who has a simulated hearing loss or when both are speaking in a noisy background. These comparisons will enable us to get a sense of the degree to which an older person is able to adapt their speech to overcome difficult listening conditions, a skill which is of paramount importance in speech communication in everyday life. We will obtain high-quality digital recordings of the participants' speech but will also, via sensors placed on the neck, record information about their vocal fold vibration, which determines the quality of their voice. Video recordings will also be analysed to investigate whether older speakers make use of eye gaze and head gestures to signal aspects of discourse such as turn-taking and back-channelling (e.g., saying 'okay' to signal understanding), to the same degree as younger speakers. In Study 2, older and younger listeners with normal and impaired hearing will be presented some of the sentence materials recorded in Study 1 by all speaker groups in good and adverse listening conditions. Tests will be presented in both auditory-alone and audiovisual conditions. Intelligibility tests will be run to see what impact age, hearing status and visual cues have on speech understanding and to see whether the 'clear speech' adaptations made by older speakers to counter the effects of poor communication conditions gives the same benefit to that of younger speakers. Sentence recall tests will also be run to investigate whether the listening effort is reduced listening to 'clear speech'. This project will lead to a better understanding of the effects of ageing on speech communication and of the various contributing factors to potentially degraded speech communication in a population of 'healthy aged' individuals. These benchmarks will be of use for practitioners such as speech and language therapists and audiologists who work on aspects of communication with older people who have health complications. A better understanding of communication difficulties that older individuals experience and of their strategies to overcome these difficulties will also assist professionals such as social workers and care professionals who work to improve quality of life for older people, as well as developers of speech technology devices for telemedicine and remote monitoring. Importantly, this research will also contribute to our basic understanding of speech perception and production development across the lifespan.

A total of 83 participants of native Southern British English adult talkers between the ages of 19 and 84 years, and their conversational partners (whose speech was not analysed) took part of the study. All participants were non-bilingual native Southern British English speakers who reported no history of speech or language impairments. The primary participants were divided into two age groups: ‘younger adults’ (YA) between 19-26 years of age (15 F, 11 M; F Mean=22 years, M Mean=21 years) and ‘older adults’ (OA) between 65-84 years of age (30 F, 27 M; F Mean=71 years, M Mean=74 years). YA participants passed a hearing screen at 25 dB HL or better at octave frequencies between 250-8000 Hz in both ears. OA participants had either ‘normal hearing’ (OANH: Female N=14; Male N=13) defined as a hearing threshold of <20 dB between octave frequencies 250-4000 Hz or a ‘mild hearing loss’ (OAHL: Female N=16; Male N=14) defined as a hearing threshold of <45 dB in this frequency range with a symmetrical downward slope of pure tone threshold in the high-frequency range typical for an age-related hearing loss profile. Informed written consent was obtained. Ethical approval was obtained from the University College London (UCL) Research Ethics Committee. During the recording, the two participants sat in different sound-treated rooms and communicated via Vic Firth headsets fitted with an Eagle G157b lapel microphones. The speech of each participant was recorded on a separate channel at a sampling rate of 44,100 Hz (16 bit) using an EMU 0404 USB audio interface and Adobe Audition and Rode NT1-A condenser microphones. In the ‘normal’ (NORM) condition, the two speakers could hear each other without difficulty. In order to elicit clear speech adaptations, in the hearing loss simulation condition (HLS) the voice of one of the talkers was processed in real time through a hearing loss simulator (HELPS; Zurek, and Desloge, 2007) mimicking the effect of severe-to-profound age-related hearing loss before being transmitted to Talker B. In the BAB-1 condition, the speech of the primary participant was mixed with the same 8-talker babble as used in the earlier adult diapix study (Hazan and Baker, 2011) before being channelled through to the confederate’s headphones, at a difficulty level equated to the HLS conditions via a Modified Rhyme Task (MRT). In the BAB-2 conditions, both participants, heard the same background babble as in BAB-1 but at an approximate level of 0 dB SNR. For all recordings, each audio channel was automatically transcribed using cloud-based speech recognition system by Speechmatics (https://www.speechmatics.com/). These automated transcriptions and the audio-transcription alignment were then hand-checked at a word level and corrected for errors to a set of transcription guidelines. Phoneme-level alignment software was used to automatically align the transcriptions and create Praat Textgrids with separate phoneme tier in addition to the word tier. Recordings lasted for about 10 minutes, yielding around 4 minutes of analysable speech for Talker A once silences, fillers, non-speech sounds such as laughter and sections with background noise had been excluded. The following acoustic measures were obtained using analyses described in the metadata information file: articulation rate, pause frequency, long-term average spectrum measures, fundamental frequency median and range, vowel space area, first and second formant ranges (see further information in metadata information file). Task transaction efficiency measures include: for diapix, the number of differences found in 10 minutes and the time taken to find the first 8 differences and for BKB repetition the percentage of correctly identified keywords by the confederate. The sensory measures collected include: pure-tone audiometric thresholds, frequency modulation and gap detection, measures of word perception in noise. The cognitive measures collected include: forward and backward digit span, word association and the Folstein mini-mental state exam.