1.1. Language input and age of acquisition in deaf infants

There are a number of specific challenges to the study of deaf infants’ perception. It is well established that, when a typically developing child is exposed to a natural and rich language input from birth, “the course of language development is largely predictable” (Lillo-Martin & Henner, 2021, p. 403). Nevertheless, early (initial) fluent input is not accessible to the majority of deaf infants, as it is well-known that most deaf children are born into non-deaf homes. Estimates based on a national survey in the United States from the turn of the millennium indicate that less than 5% of deaf and hard of hearing individuals have at least one deaf parent (Mitchell & Karchmer, 2004). Despite the newborn hearing screening programs currently in place in many countries, deafness is not always promptly diagnosed, because, for instance, it can be early acquired, after screening (Oliveira et al., 2019). Besides late diagnosis, delays in exposure to appropriate linguistic input may result from the lack of knowledge of a sign language by caregivers, who might nevertheless choose to learn sign language to communicate with their child – a process that inevitably takes time to produce the first effects; in addition, it may be difficult for caregivers to provide their child opportunities to interact with fluent sign language users early on. Often, parents do not engage in promoting sign language input to their infant; instead, they choose to take advantage of technological devices allowing the child to have access (only) to spoken language, such as cochlear implants, requiring surgical intervention usually not performed before 9 months (cf. current FDA indications for cochlear implant, at 9 months or above, Culbertson et al., 2022). Even after this intervention, there are losses in the quality of language input relative to typical hearing conditions and training is required (e.g., Lillo-Martin & Henner, 2021, for a review).

Due to such various circumstances, deaf children very often get access to (quasi-)appropriate linguistic input – that is, input in a modality that they can effectively perceive – only several months or even years after birth (Lillo-Martin & Henner, 2021; Meier, 2016). Studies with deaf children with cochlear implants show that the absence of appropriate linguistic input during the first year(s) of life causes difficulties in various areas of language development, as vocabulary growth, phonological processes or the acquisition of syntactic dependencies (Culbertson et al., 2022; Lynce et al., 2019; Moita, 2022; Nicholas & Geers, 2006; Quer & Steinbach, 2019).

It is estimated that “congenital deafness or early acquired deafness affects 1 to 3 out of 1000 newborns without risk factors and 20 to 40 out of 1000 newborns with risk factors” (Oliveira et al., 2019, p. 767). In Portugal, a survey carried out in 15 hospitals and maternities of the National Health Service found that in 2014 and 2015, 2.5 and 1.5 per 1000 newborns were diagnosed with congenital/early acquired deafness (Oliveira et al., 2019). Despite these small numbers, according to the Portuguese official channels, around 30,000 deaf people use Portuguese Sign Language (LGP) to communicate,¹ and LGP is one of the three official languages in Portugal.² Studies on LGP tend to focus on deaf signers coming from deaf families (e.g., Amaral et al., 1994; Bettencourt, 2015; Carmo et al., 2007). Nevertheless, like in other parts of the world, deaf individuals coming from deaf homes represent only a very small portion of the deaf population in Portugal, while a vast majority is born into hearing families and is not exposed to sign language early. Importantly, studies with deaf infants born into hearing homes are also relevant because they contribute to a better characterization of various aspects of the language input in language acquisition (such as the effect of input quantity, quality, and age of exposure in a perceptible modality), bearing on topics as fundamental as the critical period for language acquisition (Meier, 2016; Quer & Steinbach, 2019). Besides that, studies on this population may contribute to the development of remediation strategies to compensate the effects of delayed input, impacting health, society and individuals’ well-being.

For the reasons above, in our study on the early acquisition of prosody in LGP, we are focusing on deaf infants below 12 months of age, either from hearing homes (with early exposure to visual but not sound cues), or with early access to LGP.

1.2. Prosody in sign languages

Research in the last decades has shown that phonological organization and functioning in spoken and sign languages share core properties. In spoken languages, segmentals and their composing features (i.e., consonants and vowels, produced by action of speech articulators) are the fundamental units that constitute the form of words, and the same function is accomplished in sign languages by signed forms, produced mainly with the action of manual articulators, in particular, the dominant hand, its shape, orientation, movement and location relative to the body and head, and face areas (e.g., Fenlon et al., 2018; van der Hulst & van der Kooij, 2021; Wilbourn & Casasola, 2007). Prosody, in turn, is conveyed in spoken languages by variations in duration, intensity, pitch or frequency, marking prominence, speech chunking, sentence types and other linguistically relevant information, and there is broad consensus that the same functions are achieved in sign languages by variations in the rhythm, size and duration of gestures, pause (hold), body or head movements, and facial expressions (Brentari & Crossley, 2002; Cruz et al., 2019; Dachkovsky & Sandler, 2009; Nespor & Sandler, 1999; Sandler, 2010, and many others).

In sign language prosodic properties are conveyed mainly by facial expressions and head movement (involving for instance eyebrows, eyes, mouth, cheeks, and head position). These facial components mark prosodic constituents for various discourse functions, such as distinguishing among sentence types, like declarative sentences and wh- and yes/no questions, and marking information structure (e.g., focus/background, topic/comment, given/new; Dachkovsky & Sandler, 2009; Pfau & Quer, 2010; Sandler et al., 2020). In LGP, Cruz et al. (2019) have shown that nonmanuals distinguish sentence types: statements are typically produced without nonmanual marking; when head movements occur, the dominant one is the head (nod) up-down; in contrast, information-seeking yes-no questions are usually marked with nonmanuals, namely with eyebrow lowering together with head (nod) up-down. Scholars have also shown that nonmanual components present in sign languages cue prosodic constituents (prosodic word, phonological phrase and intonational phrase) (Brentari & Crossley, 2002; Brentari et al., 2015, and references therein; Crasborn et al., 2008; Nespor & Sandler, 1999; Sandler, 2010).

It is important to bear in mind, nevertheless, that besides prosody, nonmanuals and facial expressions in particular, can also signal other types of linguistic information, e.g., they can mark word class (e.g., adverbial and adjectival function), change sentence polarity, mark person distinctions in pronominals, or constitute obligatory components of signs’ lexical (phonological) form, in which case they are often iconic, reflecting the meaning of the sign (Pfau & Quer, 2010). Moreover, beyond language proper, facial expressions play a key role in conveying emotions, irrespective of language modality (e.g., Dachkovsky & Sandler, 2009; Elliott & Jacobs, 2013).

1.3. Early prosodic development in sign language

There is currently a knowledge gap on early prosodic development in sign language. Nevertheless, it seems clear that when deaf babies are exposed to a sign language from birth, sign language development progresses along a similar timeline as spoken language, possibly following the same developmental milestones as in spoken language (see Lillo-Martin & Henner, 2021 and Meier, 2016, for reviews). Two classical milestones, briefly described below, illustrate this parallelism.

Babbling is commonly observed in hearing babies, who start producing language-like sound sequences at around 6 months of age, increasingly more similar to the surrounding language as time proceeds, before and during the first words’ production period (e.g., Laing & Bergelson, 2020; Vihman et al., 1985). There is large consensus that babbling is a fundamental stage in language development, allowing babies to practice the articulation of the speech elements of the acquiring language, as well as promoting social and communicative interaction with caregivers (e.g., Lillo-Martin & Henner, 2021; Meier, 2016; Petitto & Marentette, 1991). Similarly, studies have shown that deaf infants exposed to gestures, around 10–14 months also produce meaningless manual gestures which reflect the components found in their ambient sign language (Pichler, 2012). Moreover, like in spoken language acquisition, it has been found that the manual babbling becomes more complex as children grow older, while deaf babies early exposed to gestures produce more complex manual babbling and a greater variety of types of manual babbling (Petitto & Marentette, 1991).

Another important developmental milestone, common to spoken and sign language, is the production of the first words/meaningful gestures. Babies’ first spoken words appear at around 10–12 months, although there is great individual variation (e.g., Lillo-Martin & Henner, 2021; Petitto & Marentette, 1991). Whereas some studies report coincidence in the time of the first emerging words and signs (between 10–12 months, Meier, 2016; Petitto & Marentette, 1991), others point to an earlier onset of the first gestures in sign language in comparison with the first emerging words in spoken language (at approximately 8.5 months), suggesting an early gesture advantage, which may persist for some time in language development (Anderson & Reilly, 2002; Lillo-Martin & Henner, 2021; see also Pichler, 2012 for a review on modality-driven differences in early language acquisition).

Most research so far has studied early prosodic development in sign language production (Brentari et al., 2015; Lillo-Martin & Henner, 2021; Petitto & Marentette, 1991). By contrast, very little is known about early perception of prosody. In fact, research on early perception abilities in sign languages has essentially looked at discrimination of various parameters (sub-lexical features) that compose signs, as well as grammatical markers. For example, Wilbourn and Casasola (2007) showed that 6- and 10-month-olds were able to discriminate the location of the sign and the signer’s facial expression cuing grammatical information, but did not demonstrate detecting changes in handshape or movement. However, to our knowledge there are no reports of research on early perception of prosody in sign language.

Interestingly, some work has addressed the early sensitivity to language-like visual contrasts by hearing infants without exposure to a sign language. A few studies have shown that hearing 3.5–4 month-old infants that are naïve to sign language are able to discriminate several contrasting parameters in a sign language, such as the type of movement (Wilbourn & Casasola, 2007, for a review). Moreover, visual attention to language-related visual information has been found to be modulated by age. Stone et al. (2017) looked at visual attention of hearing infants at the age of 6 months and 12 months while watching fingerspelling stimuli that were either well-formed or ill-formed with respect to a sonority parameter associated to syllable structure in sign languages. In this study, 6-month-olds showed a preference for the stimuli corresponding to well-formed stimuli in sign language, whereas 12-month-olds showed no evidence of preference for one type of stimulus or the other. Bosworth et al. (2022) examined the eye-gaze of hearing infants from monolingual English-speaking families, aged 6 and 11 months, while viewing video sequences of American Sign Language signs and non-linguistic body actions (self-directed grooming action and object directed pantomime). Here too, results revealed developmental differences in gaze patterns: 6-month-olds looked more to the signer’s face for grooming, and to the articulatory area of the signing space for mimes and signs; in contrast, 11-month-olds showed similar attention to the face irrespective of type of visual stimuli. Both Stone et al. (2017) and Bosworth et al. (2022) interpret the results as reflecting an early perceptual sensitivity for visual cues that can be meaningful in sign languages, followed by a later decline in discrimination abilities, before the end of the first year, associated to language specialization, analogous to what has been already established for spoken language acquisition (e.g., Werker, 2024). However, to our knowledge, visual prosodic discrimination of hearing infants naïve to sign language has not yet been investigated.

1.4. Early discrimination of intonation contrasts and the present study

For the spoken modality, Frota et al. (2014) used a modified version of the visual habituation paradigm to investigate infants’ ability to discriminate between declaratives and yes/no questions differing only in the intonation contours (falling contour vs. falling-rising contour, respectively), in European Portuguese (EP). The results showed that EP-learning infants were able to discriminate the intonation contrast between declaratives and yes/no questions as early as 5 months of age and this discrimination ability is maintained throughout the first year of life.

As far as we know, early discrimination of intonation contrasts in sign languages, including LGP, has not yet been investigated (see Lutzenberger et al., 2024 on experimental work focusing on early infant discrimination in other phonological domains, and Stone & Bosworth, 2019). In the present study, we have adapted Frota et al.’s (2014) method to investigate deaf infants’ ability to discriminate the corresponding intonation contrast in LGP. Following Cruz et al. (2019) findings, we tested infants’ discrimination of utterance-like productions with neutral expressions/absence of nonmanuals, expressing declarative meaning in LGP, and utterance-like productions articulated with nonmanuals (i.e., lowered eyebrows and head nod), used to mark yes/no questions in LGP. We expect that infants with early access to LGP will be able to discriminate intonation contrasts in the second half of the first year of life, as observed in the spoken modality (Frota et al., 2014). As for infants without or with little exposure to LGP, two outcomes may be hypothesized. Due to the limited/lack of exposure to prosodic marking in a sign language, they may show weaker/no evidence of intonation discrimination, reflecting the importance of (perceivable) linguistic input in the development of early discrimination skills. However, given the well-established early sensitivity to sound distinctions that may contrast across languages, which characterizes infant perception in general, and since hearing infants have been shown to discriminate some types of visual stimuli that contrast in sign language, we may put forth the hypothesis that infants without or with little exposure to LGP will be able to discriminate the prosodic cues conveyed by nonmanual visual cues marking the declarative/yes-no questions contrast in LGP.

2.1. Participants

Two infants took part in this pilot study: one deaf infant (male, aged 7 months and 28 days) and one monolingual hearing infant (female, 6 months and 15 days of age). Both infants were raised in EP hearing homes. The deaf infant started receiving LGP input only at the age of 6 months, for two hours, twice a week, at an educational institution specialized in early intervention for deaf infants (Agrupamento de Escolas Quinta de Marrocos). At the time of data collection, the baby had been exposed to LGP for 1 month, and received a cochlear implant 15 days before coming to the Lab.³

Informed written consent was obtained from caregivers prior to data collection. The study was conducted in accordance with the Declaration of Helsinki and the guidelines of the European Union Agency for Fundamental Rights and approved by the Ethics Committee of the School of Arts and Humanities (University of Lisbon) (4_CEI2023, 11 May, 2023).

2.2. Materials

Stimuli consisted of one-pseudosign-utterances. Eight pseudosigns were created, with the help of a native LGP adult signer, relying on already existing signs in LGP and changing one parameter (location), resulting in a possible, but non-existing sign in LGP. As yes-no questions are marked by facial cues in LGP, utmost care was taken to create pseudosigns avoiding the face region, thus articulated by manuals in body regions below the chin. This ensured a clear segregation between the regions for pseudosign articulation and intonation marking, respectively, enabling the identification of non-overlapping areas of interest (AOI).

Each of the eight pseudosigns was produced with the typical LGP statement (i.e., without facial marking) and question (i.e., with facial marking, namely eyebrow lowering, together with head (nod) up-down) intonation by a female LGP native signer. The production of the resulting sixteen one-pseudosign-utterances was videorecorded with a professional JVC camera, model GY-HM11E, in.mov format (a 4:3 aspect ratio, 25 fps). The native signer was asked to remain in the same place during the recordings, in order to allow smooth transitions in the sequences of video recordings when combined. The recorded material was edited in Shotcut software. Each recorded production of one-pseudosign-utterance was first extracted to a separate file. The 8 one-pseudosign-utterances were then combined in single sequences of 4 pseudo-sign-utterances, two sequences for statement intonation, and another two for question intonation (see supplementary materials at https://labfon.letras.ulisboa.pt/babylab/sign_language). Minor lighting and colour adjustments were made in the individual recordings, so that video sequences had similar characteristics, avoiding as much as possible exposing infants to distracting visual contrasts (Figure 1).

Figure 1

Illustration of the intonation contrast taken as still frames of the videos sequences: Statement in the left panel; Question in the right panel.

Different pseudosign sequences were used in the habituation and test phases. Each infant only saw one version (either the statement or the question) of each one-pseudosign-utterance.

2.3. Procedure

Following Frota et al.’s (2014) study, we used a modified version of the habituation paradigm, with the sound stimuli replaced by the visual stimuli. Importantly, the data were collected using the EyeLink 1000 Plus eyetracker instead of LOOK software that collects duration of a gaze, but not the location of gaze.

Infants were seated on their caregivers’ laps facing a 22-inch ASUS monitor at a distance of 50 to 60 cm. The experiment started with a short video to attract infants’ attention to the screen, while the researcher was seating at a distance, monitoring eye-movement for calibration and subsequent data collection. Parents were instructed not to interact with the infant, nor give any bodily cues that could affect infant’s response, such as pointing.

Infants were calibrated with 3 point-calibration using small moving circles as calibration points. The same procedure was used for validation. The calibration was successful in both infants with less than 0.56º of error. After calibration and validation, a coloured image (attention getter) appeared. If infants looked at it for two consecutive seconds, a video trial started. Each video trial consisted of a video file made up of sequences of 4 pseudosigns, as described above (section 2.2). The 4 pseudosigns, each corresponding to an intonational phrase, created a 16-second trial for each intonation pattern. Because our single pseudo-sign intonational phrases were longer than the single pseudoword intonational phrases in Frota et al.’s (2014), it was not possible to have both the same length (16 secs) and the same number of pseudosigns per trial matching the number of pseudowords per trial in Frota et al.’s (2014) auditory experiment (i.e., 8 pseudowords). To ensure that the total duration of the experiments across studies was similar, and, crucially, not excessively long for young infants, we decided to keep the trial length identical in the two experiments and reduced to four pseudosigns in our trials.

Infants were presented with two phases: the habituation and the test phase. In this pilot study, infants were habituated to the declarative sequence and tested with the same test order: a declarative followed by an interrogative sequence (i.e., same-switch order). In the habituation phase, infants were habituated to one sequence of 4 one-pseudosign-intonational phrases (statement intonation) until the infant reached a pre-defined habituation criterion. As in Frota et al.’s (2014) auditory experiment, we used a sliding window comparing the first four trials, with the last four trials, with the criterion that the average looking time to the last four habituation trials should be less than 60% of the average looking time to the first four habituation trials. Thus, the habituation phase had a minimum of 8 trials. The maximum number of habituation trials was set to 20. After infants were habituated to declarative intonation pattern, the test phase was presented, including two trials: one containing a trial identical to the habituation trial (same) and another different to the habituation trial (switch). Figure 2 illustrates the experimental design. If infants were able to discriminate the intonation contrast in LGP, looking time for the switch trials should be longer than for the same trials.

Figure 2

An illustration of the experimental design.

The stimuli were presented using the Experiment Builder software (version 2.3.38), and each video trial was presented until the end (fixed trial presentation). The EyeLink software was used to record and monitor infants’ looking behavior, while data extraction was done in DataViewer (version 3.2.1).

Two areas of interest (AOI) were defined to compare infants’ looking time to manuals and nonmanuals. For the manual AOI we considered the area of torso/body (from mid-neck to hips) where all hand movements were expressed. The nonmanual AOI covered the head/face region.