Somatosensory iconicity: Insights from sighted signers and blind gesturers

2.1 Iconicity through the motivation and sedimentation model (MSM)

Describing perceptual experiences employs semiotic signs (henceforth, S-signs)^[3] that may be signs, gestures, words, or pictures. According to tradition(s) of semiotics (Jakobson 1965, Sonesson 2016, 2022) an S-sign, as an expression, stands for an object, interpreted by someone, based on one or more grounds with varying predominance: iconicity (resemblance), indexicality (contiguity), and symbolicity (conventionality or habit). When something is an expression, its object is differentiated from it and is more in focus than the expression from a subjective perspective. This differentiation makes the characteristics of the S-sign phenomenologically different from those of sensory perception (Sonesson 2016, 2022). In the standard FinSL sign for ‘sharp’ (symbolicity), the first finger moves as if the knife cuts the finger of the other hand (iconicity), indicating the sharpness of the knife (indexicality); the aspect of sharpness is more in focus than the articulators themselves. Here, I will strictly focus on iconicity while acknowledging its sub-types (i.e. image, diagram, and metaphor) and the ubiquitous role of other grounds.

The condition where sensorily and linguistically different participants describe the stimuli represents a semiotically complex nexus of bodily and sociocultural factors. To manage and study this systematically, I apply a narrower version of the model that emerged within cognitive semiotics: the MSM (Zlatev and Blomberg 2019) (Table 1). This model is inspired by ideas from Coseriu’s integral linguistics (Coseriu 1985, Coseriu 2000) and phenomenology (Husserl 1901, Merleau-Ponty 1945). The MSM has been applied to study linguistic norms (Zlatev and Blomberg 2019), linguistic relativity (Blomberg and Zlatev 2021), metaphoricity (Stampoulidis et al. 2019, Devylder and Zlatev 2020, Zlatev et al. 2021, Moskaluk et al. 2022), and the origins of money (Oakley and Zlatev 2024). Unlike Coseriu’s approach, the MSM extends beyond the scope of language to encompass any systems of meaning-making and incorporates more tenets from phenomenology and cognitive linguistics. This model addresses both static and dynamic aspects, as well as individual and (potentially) more universal aspects of iconicity.

Table 1

The MSM (adapted from Devylder and Zlatev 2020, 273)

Levels
Situated		Meaning-making within an immediate situation
Sedimented		Socio-historical and symbolic conventions
Embodied		Pan-human, pre-signitivea

^aSome phenomenologists (e.g. Gallagher and Zahavi 2012) acknowledge that prior knowledge can influence perception (e.g. socially shared knowledge of the affordances of a car). Thus, the Situated and Sedimented levels may motivate the Embodied level, suggesting that the MSM may need theoretical revision. However, I leave this as a preliminary reflection outside the scope of the current study.

For the present study, the MSM distinguishes three interacting levels of meaning-making (Table 1). The situated level is the most dynamic and creative, involving on-going meaning-making within immediate contextual situations where communication occurs. The sedimented level encompasses historically, culturally, and socially shared conventions. The embodied level involves pre-signitive, perceptual, bodily, and cognitive processes and structures, such as analogy-making (Gentner and Markman 1997) and bodily schemas (i.e. individual or culture-laden habits; Donald 1998). These are pan-human features that recognise differences among populations, such as the sensory differences in the participants of the current study.

Regarding the terms motivation and sedimentation (see also Zlatev 2018), the former term, inspired by Fundierung (directly translated as foundation or grounding; Merleau-Ponty 1945, 146), describes a bidirectional relationship where a higher level of meaning-making is grounded in and sublimated by a lower level, without being reduced to one of the levels. Motivation is represented as an upward arrow (Table 1). Situated acts, such as innovative expressions, may become part of habits or conventions (somewhat comparable to symbolicity) through processes of sedimentation, shown as a downward dashed arrow (Table 1).

In the framework of semiotics, since the metaphor itself is iconic (two-object resemblance), the findings of the metaphor studies using the MSM mentioned above can apply to other types of iconicity (expression–object resemblance) relatively straightforwardly. To elaborate the iconicity using the MSM, novel iconic expressions emerge through motivation, but not determinedly, by the Embodied, especially analogy-making and perception, as shown with an upward Embodied–Situated arrow. This aligns with several SL studies on iconicity (Taub 2011, Perniss et al. 2010, Emmorey 2014).

In another direction, over the short or long term, these iconic expressions become more conventionalised and thus sedimented into interpersonal or communal norms, as depicted by a downward Situated–Sedimented arrow (Table 1), in parallel with several scholars mentioned above. As is well known, several lexical signs originate from spontaneous ones, often called gestures or pantomimes (Taub 2011, Ortega and Özyürek 2020b).

Signs do not solely become fully conventional through the process of lexicalisation (i.e. sedimentation), as some linguists have argued (e.g. Frishberg 1975). Lexical signs can also ‘come back to life’ through the process of de-lexicalisation (Cormier et al. 2012) or re-iconisation. According to the MSM, this illustrates how an expression can be doubly motivated by the levels of Sedimented and Embodied. To elaborate, the ratio of sedimentedness/motivatedness of an expression can vary in a situated context, as roughly categorised by Johnston (2013) into three groups: a) non-conventional, b) partly conventional, and c) fully conventional. ^[4] These groups refer to a) spontaneous, gradient iconic expressions where the iconicity is vivid and immediate, b) highly iconic signs that have some conventional features (e.g. handshape), and c) non-gradient signs – those typically listed in dictionaries – whose iconicity is not immediately vivid at the moment of production, whether they were originally arbitrary or iconic (i.e., de-iconised), respectively (see a footnote for clarification).^[5]

Importantly, the Situated level, where one describes to another, involves intersubjectivity in many senses. For instance, individuals adjust their articulations to ensure they are suitably perceived by others (e.g. a sighted person, a deafblind person, or a large audience), balancing articulatory ease and perceptual clarity (Cutler et al. 1987, Emmorey 2005), or ensuring intelligibility for non-signers by gesturing slowly (Moriarty and Kusters 2021). The use (or lack thereof) of iconic descriptions in interactions can imply social power dynamics (e.g. acceptability, associations with certain linguistic communities, or social motivations for the choice of semiotic strategies; Hodge and Ferrara 2022). In the current study, the social dimension may manifest in the participant–researcher relationship within a certain research design (e.g. encouraging blind participants to use an unaccustomed gestural modality to produce intelligible iconic expressions for a sighted person).

As is known, iconicity is closely intertwined with indexicality, which is based on contiguity (e.g. Sonesson 2014, Keränen 2023). In this study, it is worth mentioning a type of abductive index: prior knowledge serves as a condition for establishing contiguity between an S-sign and an object. For example, an iconic hand that resembles holding indicates something being held (Keränen 2023; see Section 2.4).

2.2 Iconic strategies

Both SL and gesture share relatively universal types of iconic strategies (also known as depicting signs, classifiers, modes of representation, and so on). These iconic strategies have been categorised into two or more types with different labels and taxonomies (e.g. Schembri 2003, Hassemer 2016, 121).

Here, I categorise 12 main groups of iconic strategies (Figure 1, Table 2) based on their prototypical conceptualisations evoked by gestural articulators, drawing from existing literature and findings from the current study. In practice, if a certain strategy does not fit into any existing type, I either establish a new type for this finding or classify it as uncategorised (see Section 3.4). This categorisation is sensitive to unique iconic conceptualisations and potentially expands the taxonomy of iconic strategies. It differs from more traditional approaches that strictly adhere to only a few types. To briefly explain Figure 1, an arrow depicts motion, and the symbol ‘ø’ depicts non-motion.

Figure 1

Illustrative examples of types of iconic strategies (a–l).

The types that form the most fundamental iconic conceptualisations are acting (articulators ‘miming’ themselves in bodily action) and representing (articulators ‘miming’ entities other than themselves) (Müller 2014, 1696). In acting, the hands and body move as if performing actions with or without an object (Cormier et al. 2012, Keränen 2023). This strategy can also involve the entire body to convey actions, thoughts, and emotions (constructed action, e.g. Cormier et al. 2015). In representing, a hand resembles entities other than itself, and the whole body can be represented as a non-human entity (e.g. an animal; also known as personification) (Hwang et al. 2017). Moreover, the acting and representing strategies can involve different body parts at the same time (see body partitioning by Dudis 2004, Cormier et al. 2012, 344), such as using an arm to represent a human arm and a first finger to represent a toothbrush, as seen in the type known as instrument (Padden et al. 2013).

There are other strategy types, but they may be emergent from the two fundamental iconic strategies, according to Müller (2014). For example, gesture drawing, or tracing, in the air may originate from the hand drawing on the surface. Mostly acknowledged types in SL linguistics are tracing (e.g. drawing and moulding by Müller 2014) and measuring (Mandel 1977, 69, Hassemer 2016, Hassemer and Winter 2016), which are often classified under the type Size and Shape Specifier, or SASS (e.g. Johnston and Schembri 2007, 170). Less acknowledged ones are locating (Liddell 2003), point of reference (POR) (Johnston 2019), static outlining, or just outlining (Calbris 1990, holding by Hassemer 2016), and being-pulled (Keränen 2023). Additionally, the types found in the current study are dividing, emptying, and assembling.

To explain each type, in tracing, a hand or some body part moves to dynamically ‘draw’ shapes in the air or on the surface; in measuring, two or more fingers or hands are opposite to each other to measure the distance between them. In locating, a hand uses a short movement in a particular direction, as if placing something on the ground to show its location. In POR, a non-dominant hand represents itself as a less or more abstract background (ground) to show the spatial relation between it and an active object (figure). In outlining, a hand or two hands with certain configurations stay in the air to show different geometric shapes (e.g. a circle or heart) using static handshapes and/or arms. In being-pulled, a hand pulls something to show that something is absorbed, magnetised, collected, and so on. In dividing, which is formally close to tracing, a hand moves over something to divide, cut, or split it into two pieces, instead of tracing shapes. In emptying, when a non-dominant hand holds an imaginary hollow object, a dominant one usually puts itself into the hollow object to show the spatial emptiness inside it.

Then, assembling strategy may be regarded as an intermediate form between the paradigm (i.e., selection of strategies) and the syntagm (i.e., combination of strategies). In this type, it is like assembling parts into a whole object in a tracing-like but discontinuous manner. For example, in the sign SQUARE, a signer quickly repeats a pinch-like L-handshape (i.e. measuring a certain length) at two different angles (e.g. vertical and horizontal lengths) to create a complete object (e.g. a square formed by the two lengths at different angles). Unlike dynamic tracing (e.g. Figure 1e), assembling leaves traces after each repetition of a non-dynamic strategy (e.g. measuring). Unlike the simple combination of different strategies (e.g. acting and tracing), assembling forms a unified whole object, similar to tracing.

The boundary between fundamental strategies and others may occasionally be fuzzy. For example, a signer can use either the acting or tracing strategy with the first finger to draw shapes. However, the most important difference between the acting for drawing and the tracing for drawing lies in the question of which aspect is in focus from a subjective view.^[6] In the former, the actor’s body itself – sometimes also the shape – is highlighted, and in the latter, only the shape is highlighted.

Moreover, I also preliminarily categorise the iconic substrategies of each main type, and each substrategy may be either acting or representing – possibly also indexing (i.e. pointing finger; see Section 4.2.1). The main strategies are responsible for primary iconic conceptualisations, as shown above, but are somehow influenced or co-construed – lacking a better term – by different substrategies to create slightly different but unified iconic conceptualisations. A substrategy can often be identified by observing handshape. For example, a clock can be located on a wall with either sub-handling (i.e. grasp-like C-handshape; Figure 1g) or sub-representing (i.e. a hand as a clock; Figure 1h). More examples are found in Table 2. To speak metaphorically, the relationship between the main strategy and substrategy is like painting a picture (main strategy) with different instruments, such as a finger, a brush, or a roller (substrategies), in parallel to the notion of ‘depictive techniques’ by Müller (2014). Note that the grasp-like handshape in the non-acting strategies has also been mentioned elsewhere, albeit with little attention given to it (e.g. Johnston and Schembri 2007, 171, Perniss and Vigliocco 2014, 2–3, Hassemer and Winter 2016, 407–8).

2.3 Tendencies for the use of iconic strategies

Studies have shown that there may be shared and different tendencies across people due to various factors. Sighted signers and gesturers share a tendency to use certain iconic strategies for certain semantic fields consistently. This is commonly explained by the shared bodily and visual affordances (Padden et al. 2013, Ortega and Özyürek 2020b, Keränen 2021) – due to Embodied motivation. According to them, the acting strategy is most frequent, especially when describing bodily actions; the tracing strategy is generally less frequent but more frequent when describing the shape of an object or a non-manipulatable object (e.g. a house). Moreover, in both gesture and SL, two hands can recruit different iconic strategies simultaneously (Hwang et al. 2017, Slonimska 2022) or sequentially (Ortega and Özyürek 2020a) to express a concept efficiently. Then, regarding differences, Padden et al. (2013) show that when describing a handling action (e.g. toothbrushing), signers may tend to use the instrument strategy (an arm as a human arm and a hand as a toothbrush), and gesturers tend to use the acting strategy.

Regarding Embodied–Situated motivation, the understanding of how different sensory capabilities (e.g. deaf, hearing, and blind) influence iconicity or gestural expressions is still premature. First, congenitally blind children were reported to use notably fewer gestures but more exclusively speech in the Directions Task (describing paths, landmarks, and locations) compared to sighted children (Iverson and Goldin-Meadow 1997). They concluded that sighted and blind children rely on global (i.e. vision globally navigates a path) and segmented (i.e. a path broken into landmarks) representations, respectively; the latter representation is less suited to producing gestures. Second, the fact that blind signers cannot see another’s facial expressions can lead to disappearance (Checchetto et al. 2018, 2) or even the preferable avoidance of using facial expressions (Edwards and Brentari 2020). Third, the lack of visual experience can affect the organisation of knowledge, resulting in different frequencies of iconic strategies across the groups of blind and sighted gesturers (Mamus et al. 2024).

Regarding Sedimented–Situated motivation, Özçalışkan et al. (2016) reported that, in co-speech gestures, congenitally blind and sighted participants of the same language share language-specific gesture patterns but differ from those who speak a different language. That is, blind participants adhere to language-specific co-gesture patterns, even though they have never seen cultural gestures before. However, the language-specific patterns do not affect any groups when using only silent gestures, suggesting that silent gestures follow a natural semantic organisation rather than being influenced by vision and language (Özçalışkan et al. 2018). For this reason, the current study focuses on signing and silent gesturing without incorporating cross-linguistic comparison.

2.4 Embodied motivations by exploratory procedures (EPs) and grasps

Since iconic S-signs are ultimately motivated by the sensorimotor (and emotional) system (e.g. Perniss and Vigliocco 2014, Zlatev et al. 2008, Zlatev 2018), I now elaborate on iconic motivation by retrieving a few findings from the studies related to cognition and the body.

A physical object has many properties, such as texture, size, shape, and weight. According to Lederman and Klatzky (1987), human hands purposively move and touch the object in a patterned manner to optimally gain desired information about its haptic properties; they classified these prototypical movement patterns as types of exploratory procedures, or Eps (Table 3). Also, hands may haptically explore the functions of the object (e.g. pushing a press) (ibid).^[7]

Moreover, reaching and grasping come in different manners in contextual sensitivity to a given task (i.e. goal), and the physical properties of an object (e.g. Ansuini et al. 2006). Grasps may be categorised into three main types: precision, intermediate, and power (Feix et al. 2016): prototypically, the heavier or bigger the object is, the more fingers and palms are recruited to handle it. While pinching a needle recruits the two fingertips of the first finger and thumb (precision), the act of effortful hammering a nail recruits all fingers and palms (power grasp); the intermediate grasp combines both elements of precision and power grasps, as in rotating a key (thumb on the first finger). These show the hand–goal–object interdependency: manual movements and positions correlate with the goal and the properties of an object.

Regarding Embodied–Situated motivation, it is easy to find that many gestures and signs are motivated by EPs. For example, the FinSL sign for ‘soft’ – pressuring an object with ten fingers – is iconic for the pressure type of EPs. Moreover, an iconic grasp-like handshape (i.e. handling substrategy) abductively indicates the object being held, thanks to the motivation by hand–goal–object interdependency (see the embodied nature of gesture by Hassemer 2016).

Furthermore, Prinz (2013, xi) puts it in the foreword: “Upon seeing one surface of a ball, we may use tactile imagery to image its sphericality.” Thus, vision tends to evoke haptic imagery without actually touching it. To apply this, seeing the surface of the hands presupposes the haptic aspects of them. This seems to enable haptic iconicity (see Section 4.2.2).

In sum, based on the literature and phenomenological reflections, since blind and sighted people share bodily senses and functions, they are expected to share iconic strategies. On the other hand, blind (silent) gesturers’ iconicity is expected to be motivated by their sensorimotor experience, thus proprioception (Embodied) rather than SL conventions (Sedimented), due to the lack of SL skills. Thus, the groups may differ depending on the processes of motivations and sedimentations, including linguistic, sensory, and intersubjective dimensions, as shown in the MSM (Section 2.1). Now, I proceed from the first-person conceptual analysis to the third-person experiment.

3.1 Participants

The selected groups of legal adult participants (male = 57%; female = 43%; average age = 45.7) were 5 sighted deaf L1 signers (SS1–5) and 2 congenitally blind gesturers who hear (BG1–2). The participants were invited either directly or indirectly through organisations with general information about the study. Participants’ information was collected through the pre-task questionnaire in place. Communication with signers was conducted in FinSL throughout the sessions, and with gesturers through FinSL interpreters – used primarily during pre-task communication and after-task interviews – who were prepared using my advance materials.

To provide context, in Finland, with a population of approximately 5.5 million, there are about 3,000 deaf signers (The Finnish Association of the Deaf, n.d.). According to the Finnish Register of Visual Impairment Annual Statistics 2021, 22% of the 55,000 visually impaired people are blind, based on the definition by the World Health Organisation that their activities primarily rely on senses other than sight (Tolkkinen 2022).

Regarding the blind participants in this study, the first blind gesturer is congenitally blind with had very poor sight until adolescence, primarily perceiving objects close to the face with blurry vision. Gesturer BG2 has only limited light vision. While he could not see any surface at all or say whether lights were on in the room where tasks were conducted, he could sometimes experience some brightness (e.g. sun). Therefore, we must interpret the results from gesturer BG1 with caution because he has visual memories that may influence the formation of iconicity. It is known that human vision and visual experiences develop significantly during the first months of life; the ability to perceive the three-dimensional world is typically reached by 6 months of age (see a review by Siu and Murphy 2018).

3.2 Apparatus

The sets of stimuli are physical, three-dimensional objects. All of these are haptically accessible to the group of blind participants and visually accessible to the other group. In addition, a benefit of using physical objects as stimuli is that all participants describe the exact same objects. In contrast, for example, a word list can elicit various objects based on participants’ denotative and connotative associations. Thus, using physical objects as stimuli makes descriptions between the groups comparable. The sets of stimuli are divided into two groups (a and b) based on Tasks 1 and 2, respectively:

1. 18 common manual household items that are typically used with hands;

2. 12 pairs of geometric objects, ranging from more typical (e.g. sphere, cone, etc.) to less typical (e.g. a pentagonal pen holder).

I expect that a) the set of household items with familiar manual functions will elicit the acting strategy, and b) the set of geometric objects will elicit shape- and size-related strategies. In the latter of the set, the pairs of objects are expected to motivate comparison and therefore more detailed iconic descriptions of them.

3.3 Procedure

In both Tasks 1 and 2, participants (Figure 2a) from both groups were seated on a chair, with the sets of stimuli located on a table to their left (Figure 2b), taken from behind a visual barrier (Figure 2c). Participants were instructed to 1) briefly observe each stimulus on the table (about 5 s) before moving it out of sight behind another visual barrier (Figure 2d),^[8] and then 2) describe it to me, a sighted deaf person (Figure 2e), encouraging the use of intelligible expressions. They were asked to provide concise yet detailed descriptions so that any deaf person could conceive of the objects without seeing them. If needed, the participants were asked – through FinSL or via the interpreters – to describe in more detail. In Task 1, the participants were asked to describe what the household items are and to show how they are used. In Task 2, they were asked to describe the size and shape of the pairs of geometric objects. After the two task scenarios, I asked debriefing questions to the participants, to attempt to understand something about their descriptions (Section 4.2.4), either through FinSL or via the interpreters. The three sessions were recorded using two video cameras (Figure 2f), and then annotated and analysed with the time-aligned video annotation software ELAN (Crasborn and Sloetjes 2008).

Figure 2

Illustration of the session set-up. (a) participant, (b) stimuli presented on the table, (c) stimuli hidden under the table before presentation, (d) stimuli re-hidden behind the barrier, (e) a sighted deaf researcher (myself), and (f) two cameras.

The set-ups were slightly different according to the groups. The groups of sighted signers and blind gesturers were asked to only see and only touch the stimuli, respectively. This ensured that the results of the sighted signers were not based on fresh haptic memories of the stimuli. Also, an interpreter was seated at the back of each blind participant, near the table.

3.4 Annotation

Applying and modifying the annotation conventions established in the corpus project of FinSL (Salonen et al. 2019), the annotation procedure was done in the general-specific loop shown below to systematise the data and minimise the number of errors. Exceptionally, with the blind gesturers, the annotation started at the second point due to the lack of lexical signs. In the annotation template (Figure 3), the tiers are as follows: ID (i.e., the tier for form-based glosses), Right hand, Left hand, Both hands (as a single unit), Substrategies as child tiers of the hand tiers, Groups of stimuli (grouping tokens for each stimulus), and personal notes made by me. For now, the last two tiers function as support for my procedural work. The list of ID glosses comes from the controlled vocabulary of ELAN, which is connected to and updated by the lexical database of the Finnish Signbank. Moreover, I also established my own the offline-controlled vocabulary for annotating specific semiotic types.

1. While annotating all signs at the tier ID, I labelled fully conventional signs with ID gloss, partly or fully iconic with a general iconic gloss (i.e., instead of specifying subtypes), indices with a general index gloss, and unidentifiable tokens with an unidentifiable gloss. In doing so, it was possible to distinguish iconic tokens from others systematically.

2. After that, at the separate tiers of hands (left and right), fully conventional tokens (see Section 2.1) were labelled as non-iconic (i.e. not evoking iconicity); indices remained as general indices; and iconic tokens were labelled according to the types of iconic strategies. Then, iconic tokens whose iconic strategy was unclear were labelled as uncategorised for further rechecking and for the possibility of creating new types for them. More concretely, unclear iconic strategies are those that, for example, do not fit any existing categories or are simply ambiguous.

3. The third tier for both hands was created by copying the annotations from both hand tiers. This third tier describes hands as a single unit (i.e., regardless of the different strategies used by both hands) to make descriptive statistics sensible. Single units were typically annotated according to the dominant hand that often contributes more significantly to forming the iconic strategy (e.g. tracing versus POR) and tend to exhibit a higher frequency of production.

4. At the child tiers of the hand tiers, the iconic tokens were labelled according to types of substrategies: handling, representing, or uncategorised.

5. With the help of the search systems of ELAN, I systematically checked the annotation consistency across tiers to find possible errors (e.g. empty or exceptional annotations in the search results).

6. While thoroughly rechecking the uncategorised iconic strategies in the annotated data, a procedure loop is conducted (2–4).

Figure 3

Illustration of tier dependencies in the ELAN template.

4.1 Descriptive statistics

The descriptive statistics presented below provide familiarity with the results and support the arguments outlined in Section 4. For clarity, frequency tables are included in Appendices A (covering all semiotic types) and B (specifically focusing on iconic substrategies). While Appendix A encompasses all semiotic types, Sections 4.2 and 4.3 strictly focus on iconic (sub)strategies, with numeric results based on these categories.

Generally, the results reveal both similarities and differences across the participants and groups. To provide an overview, the total production frequency – regardless of handedness, strategies, and tasks – amounts to 3,242 tokens in the group of sighted signers (avg. 648.4 per signer) and 252 tokens in the group of blind gesturers (avg. 126 per gesturer). Unsurprisingly, the frequency of the fully conventional tokens is 61.9% of all tokens in the group of signers and 0% in the group of gesturers. As mentioned in Section 2.1, most iconic signs are gradually between non-conventional and partly conventional. However, the consideration of conventionality, including indexicality and unidentifiable cases, is outside the scope of the present study.

4.1.1 Quantising main iconic strategies

Turning to the consideration of iconic strategies, as expected, Task 1 tends to involve strategies that depict actions or functionalities related to the items, while Task 2 focuses more on strategies that depict the shape and size of geometric objects. In addition, exceptions were observed. For example, in Task 2, the acting strategy was used unexpectedly. For instance, a stimulus like an ice cream stick elicited the acting strategy to represent its commonly recognised specific shape. Additionally, because both groups typically combined strategies sequentially (Ortega and Özyürek 2020a), the frequencies of strategies are not mutually independent. However, this aspect is beyond the scope of the present study. It is important to note that not all participants used all strategy types; rather, there was variation in how they employed different strategies.

To consider the sighted group in Task 1 (as illustrated in Figure 4), the average frequency and percentages (a specific iconic strategy per total number of iconic strategies) for each type illustrate the most–least predominance of the types, as follows: acting (31.2; 42.6%), tracing (20; 27.3%), instrument (7.2; 9.8%), representing (7; 9.6%), measuring (6; 8.2%), being-pulled (0.6; 0.8%), assembling (0.6; 0.8%), locating (0.4; 0.5%), dividing (0.2; 0.3%), and no tokens shown in the rest.

Figure 4

Frequency diagram of iconic strategy types across sighted signers.

Task 2 shows the different predominance order, as follows (Figure 4): tracing (68.6; 65%), measuring (21.8; 20.6%), outlining (5.8; 5.5%), acting (4.4; 4.2%), assembling (1.8; 1.7%), representing (0.8; 0.8%), locating (1; 0.9%), emptying (0.8; 0.8%), dividing (0.4; 0.4%), instrument (0.2; 0.2%), and no tokens shown in the rest.

To consider the blind group in Task 1 (Figure 5), since BG1 and BG2 gesturers greatly varied in the overall production frequency and the sample size of the group was small, it is more illustrative to describe them separately. The predominance order of iconic strategies is roughly as follows: acting (BG1: 30; 30.6%; BG2: 26; 78.8%), tracing (BG1: 48; 49%; BG2: 2; 6.1%), and representing (BG1: 2; 2%; BG2: 2; 6.1%) strategies. In sum, while B1 uses slightly more tracing over acting, BG2 uses overwhelmingly more acting over others.

Figure 5

Frequency diagram of iconic strategy types across blind gesturers.

Again, Task 2 shows the different predominance order as follows (Figure 5): tracing (BG1: 52; 78.8%; BG2: 28; 73.7%), measuring (BG1: 14; 21.2%; BG2: 3; 7.9%), outlining (BG1: 0; 0%; BG2: 5; 13.2%), representing (BG1: 0; 0%; BG2: 2; 5.3%), and no tokens shown in the rest. Thus, they resemble each other in the percentage of the use of tracing but slightly differ in other types.

Appendix A – which is based on single whole units (Section 3.4) – does not show the frequency of the POR strategy because it occurs only as the subordinate role, mostly in a non-dominant hand. However, when considering the POR strategy in the left-hand tier using the ELAN search, the results show 191 tokens in the sighted group and 7 tokens in the blind group.

In conclusion, I present two points. First, the different frequencies, especially of the overall tokens and the diversity of iconic strategies, between the two groups can be partially explained by the presence of different semiotic repertories across the groups, such as the habitual use of bodily expressions and linguistic skills in the sighted group.

Second, the shared strategies in both groups were found: acting, tracing, representing, and POR in Task 1, as well as tracing, measuring, outlining, representing, and POR in Task 2. Importantly, BG2 with light vision used all these shared strategies. Thus, it can be safely concluded that since the blind gesturers do not share visual and linguistic aspects with sighted signers, the evidence of the shared strategies (i.e. acting, tracing, measuring, outlining, representing, and POR) supports the assumption of the role of proprioception in iconicity in both groups.

4.2 Qualitative descriptions

While both groups purposively use iconic bodily expressions, the surprisingly salient difference between them is the articulatory difference. Both blind gesturers use less diverse handshapes, often using a somewhat curled, loose handshape. In this handshape, the thumb, first, and middle fingers touch each. Interestingly, this is in contrast to the findings by Iverson and Goldin-Meadow (1997) according to which blind and sighted children gesture in similar ways, including motions and handshapes. To speculate, the articulatory difference may arise from the non-habitual use of bodily expressions in adult blind gesturers or from individual variation by change. Despite the lack of explanation, it may be useful to bear the articulatory difference in mind throughout the text. The following sections consider more specific types of iconic strategies.

4.2.1 Grasp-tracing and line-tracing

Surprisingly, the qualitative and quantitative use of tracing differs between the two groups, especially in the selection of substrategies. To clarify, grasp-tracing involves mimicking the action of grasping a cylindrical object (i.e. handling substrategy) and moving upward to outline its height (Figure 6a). On the other hand, line-tracing refers to extending a line in the air, possibly using the first finger, to form shapes (Figure 6b), although it remains an uncategorised substrategy.

Figure 6

Two kinds of tracing are: (a) grasp-tracing and (b) line-tracing.

The results from both Tasks 1 and 2 show that while the sighted group (SS1–5) tended to utilise a more ‘grasp-tracing’ strategy (57.3% of all tracing tokens), the blind group (BG1–2) utilised it less (only 27.6% of those) (Table 5). Instead, the latter tended to utilise the ‘line-tracing’ strategy. Importantly, since the line-tracing can be equally interpreted as acting (e.g. finger-drawing), representing (e.g. pencil), or index (finger-pointing), termed as pointed implement by Mandel (1977, 67), it was grouped into the uncategorised type (39.4% for SS1–5; 72.4% for BG1–2).

Table 5

Frequencies and percentages of substrategy types with the main tracing strategy

	Frequency		Percentage
Tracing + substrategies	SS1–5	BG1–2	SS1–5	BG1–2
Handling	246	32	57.3%	27.6%
Uncategorised	169	84	39.4%	72.4%
Representing	14	0	3.3%	0.0%
Total	429	116	100.0%	100.0%

The significant difference between the two substrategies is their ways of highlighting the specific aspect of an object. Unlike the first finger, the grasp-like handshape includes advanced information on three-dimensional properties (Hassemer and Winter 2016). As observed in my data, while the grasp-like handshape is sufficient to trace a three-dimensional thick ring with one or a few movements (Figure 6a), the first finger requires more repetition to achieve a similar outcome. For example, gesturer BG1 moved their first fingers through opposite curved paths with a repeated spiral motion to trace the thick ring (Figure 6b).

In this context, the advanced semiotic repertoire of signers potentially explains the selection of these substrategies. However, the role of vision in this selection remains unclear. One possible explanation for the increased use of line-tracing in the blind group could be that their tracing is more strongly motivated by specific EP, such as contour following (Section 2.4).

Finally, although the blind group notably used the grasp-tracing substrategy less frequently, they still employed it. For instance, gesturer BG2 traced a thick ring with a C-handshape, which effectively mimics grasping the ring. This demonstrates a clear motivation influenced by the power grasp (Section 2.4) and suggests that proprioception indeed plays a role in the formation of at least the handling substrategy.

4.2.3 Sensory and semiotic ratios in articulation

After the previous section, one might enquire about the role of vision in iconicity. As mentioned earlier, articulation in signers depends on both visual and proprioceptive feedback (Emmorey et al. 2009). I now argue that both types of feedback likely contribute to the formation of iconicity in signers, as evidenced by the outlining strategy observed exclusively in Task 2. In this strategy, hands are configured into a static shape to resemble the shape and size of the object.

Regarding the blind group, only gesturer BG2, who has some light vision, used this outlining strategy. In all five instances where outlining was used, his handshape closely resembled the literal grasp of the object (i.e. handling substrategy). When BG2 described the palm-sized, roof-shaped object, he started with the line-tracing strategy to draw its contour and continued with the grasp-like outlining strategy to iconically show its palm-sized property (Figure 7a) while saying aloud ‘size’.^[9] To conclude, the blind gesturer’s handshape must be motivated by a certain grasp type (see Section 2.4).

Figure 7

Different kinds of outlining. (a) simple grasp-like handshape, (b) rounded pinch-like F-handshape, (c) two joint grasp-like C-handshapes, and (d) oval shape shown by two curved arms.

Whereas the outlining strategy is similar across the two groups in many aspects, they also differ from each other. In the first of three examples, the sighted signers used a simple grasp-like handshape identical to the one used by BG2 (Figure 7a). Second, their handshapes are not visually identical to the simple grasp in many cases, but they also form visually sharp figures, for example, showing a small circle with a rounded F-handshape (Figure 7b) or a big circle with two joint C-handshapes (Figure 7c). While F- and C-handshapes resemble precision and power grasps, respectively (Section 2.4), the contours of curvy palms in these handshapes visually look like a clear circle. Third, a handshape does not resemble the grasping hand at all. When outlining the oval-shaped lid of the box, signer SS2 positioned his two hands in a way that the space between them visually resembled the oval shape (Figure 7c). Remarkably, the signers occasionally gazed at their hands (i.e., visual feedback) to adjust them and create visually sharp figures. In sum, there are four kinds of outlining (Figure 7a–d).

In sum, it can be safely concluded that there is a gradual ratio between proprioception and visual iconicity. On the one hand, the sighted signers outline shapes, relying on the varying degrees of two sensory feedback, ranging from more proprioceptive iconicity (simple grasp) to less or no proprioceptive iconicity (curved arms). On the other hand, congenitally blind gesturers probably rely solely on proprioceptive feedback and thus iconicity. This finding reminds us of Mitchell’s (2005) critique of the notion of ‘pure’ visual media as inexact and misleading. Instead, he argues that there is a sensory/semiotic ratio: visual media is mixed media, including other senses and semiotic grounds. While his emphasis was on the observer’s perspective (i.e. simply looking at media), my articulatory perspective (i.e. producing it) provides a complementary insight into Mitchell’s concept.

This conclusion also questions the dualist notion of sensory iconicity. In this notion (e.g. (Mesch et al. 2015), tactile iconicity and visual iconicity are classified exclusively for blind and sighted individuals, respectively. My findings support a non-dualist, gradual notion of visual and proprioceptive iconicity, at least for sighted signers.

4.3 Brief reflections on limitations

While the current study has been productive, there are limitations. First, due to the space limit, the current study exclusively focused on manual iconic strategies, however, acknowledging that both groups have also used non-manual articulators (e.g. bodily reenactments, and iconic mouthings). Their contributions to (somatosensory) iconicity, both independently and in relation to manual articulators, deserve greater attention in the future.

Second, here the categorisation of iconic strategies is phenomenologically sensitive to different iconic conceptualisations, extending the taxonomy, rather than bundling some types into one and missing their uniqueness. However, the taxonomy is still incomplete. There are some unclear cases of how they should be categorised, partially due to the fuzzy boundary between the types, as well as the sample size of the study. Moreover, while the categorisation of substrategies provided a detailed explanation, especially for the handling of substrategy, a large portion of the substrategies remained uncategorised due to their ambiguity (see also Section 4.2.1). On the other hand, such sensitive categorisation is the right way forward, as it provides a nuanced understanding of the diverse and dynamic nature of iconicity, despite the challenge of handling difficult and fuzzy cases. It is important to remember that the task of qualitative research is not only to deepen the understanding of phenomena but also to accommodate and tolerate the complexity inherent in these phenomena (Juhila 2021).

Third, the elicitation method in the current study is still quite visuocentric. In the tasks, blind gesturers are asked to describe to a deaf-sighted individual, likely adjusting their descriptions to ‘fit’ a sighted individual’s understanding. We have not yet explored alternative set-ups in which participants with different sensory and linguistic resources are combined in various ways. In fact, a blind–blind situation with physical contact can enable unique co-formed patterns (Mesch et al. 2015, Edwards and Brentari 2020). The four hands and legs of two deafblind signers can be ‘combined’ into an iconic whole (e.g. interlocutor’s hand as a tree). Exploring different set-ups could provide more nuanced and convergent findings, helping to untangle the complex interactions of sensory, linguistic, and intersubjective factors.

Fourth, the small sample, consisting of five signers and two blind gesturers, limits the ability to provide explanatory statistics and hinders generalisation. Partly due to Finland’s situation, which stems from its relatively small population (5.5 million), it is challenging to recruit a diverse range of voluntary participants, particularly those with disabilities. To address this issue, more international studies are needed to obtain a larger sample size. However, the qualitative, comparative approach has allowed for a more nuanced understanding of iconicity. Additionally, the current results can be somewhat generalised, as they align with findings from several studies on similar topics.