Recent evidence suggests that preverbal infants' gaze following can be triggered only if an actor... more Recent evidence suggests that preverbal infants' gaze following can be triggered only if an actor's head turn is preceded by the expression of communicative intent [1]. Such connectedness between ostensive and referential signals may be uniquely human, enabling infants to effectively respond to referential communication directed to them. In the light of increasing evidence of dogs' social communicative skills [2], an intriguing question is whether dogs' responsiveness to human directional gestures [3] is associated with the situational context in an infant-like manner. Borrowing a method used in infant studies [1], dogs watched video presentations of a human actor turning toward one of two objects, and their eye-gaze patterns were recorded with an eye tracker. Results show a higher tendency of gaze following in dogs when the human's head turning was preceded by the expression of communicative intent (direct gaze, addressing). This is the first evidence to show that (1) eye-tracking techniques can be used for studying dogs' social skills and (2) the exploitation of human gaze cues depends on the communicatively relevant pattern of ostensive and referential signals in dogs. Our findings give further support to the existence of a functionally infant-analog social competence in this species. Results Although recent research has provided important evidence about dogs' social communicative skills [2], it is still unclear whether dogs' gaze following is tuned to cues that signal the human's communicative intent (e.g., eye contact, verbal addressing). In order to investigate this first, we have collected eye-gaze data from 16 adult, task-naive pet dogs. Subjects were presented with a series of movies in which a human female turned her attention toward one of two identical containers either in an ostensive-communicative (O) or in a nonos-tensive (NO) manner. In the ostensive condition, the human actor overtly expressed her communicative intent, whereas in the nonostensive condition, we removed the ostensive signal from the stimulus. Each trial consisted of introductory, addressing, and cueing phases (Figures 1A–1C). All dogs participated in both O and NO trials that were different only in the addressing phase. We obtained valid data for analysis from 13 dogs in the ostensive and 14 dogs in the nonostensive condition; however, only 11 of them provided valid data in each of these conditions (for the validity criteria see Data Analysis). In the addressing phase, dogs spent similar amounts of time gazing toward the human actor in the two conditions (mean 6 SEM: 1,088.8 6 181.1 ms in O and 980.9 6 267.8 ms in NO conditions , ns) and invested a comparable amount of time scanning the region containing the actor's face relative to the whole body: mean 6 SEM = 0.46 6 0.09 in the O condition and 0.55 6 0.10 in the NO condition [paired t test t(10) = 20.88; p = 0.39] showing that in the addressing phase, the human actor evoked the same level of visual attention in both conditions. Next we analyzed whether dogs looked longer at the gaze-congruent area (Figure 1D) as compared to the gaze-incon-gruent area (cumulative accuracy). In accordance with infant eye-tracking studies (e.g., [1]), difference scores were calculated for this variable. We found that subjects looked longer to the gaze-congruent area than to the gaze-incongruent area after having seen ostensive addressing (one sample t test, t(12) = 2.382; p = 0.034). However, this was not the case for the nonostensive condition in which the difference score did not differ from zero [one-sample t test, t(13) = 20.756; p = 0.46], indicating no tendency to follow the human's gaze in the absence of communicative addressing (Figure 2). This differential sensitivity to human referential gestures is strikingly similar to that found in a study of 6.5-month-old human infants [1]. A similar analysis on dogs' first look did not reveal any significant bias toward the gaze-congruent area in O or in the NO conditions [one-sample t test, t(12) = 1.167; p = 0.266; t(13) = 20.105; p = 0.91, Figure 2]. A within subject analysis of the difference scores for cumulative looking time in the two experimental conditions was run on the 11 subjects that gave valid data in both conditions. This analysis shows that dogs were more likely to follow the model's gaze in a gaze-congruent manner in the O than in the NO condition [t(10) = 2.49; p = 0.03; the effect was independent of presentation order; see Supplemental Results available online]. However, no difference was found between conditions for the first look [t(10) = 21.21; p = 0.25]. We also explored the spatiotemporal pattern of eye movements during cueing phase, investigating how the gaze points move away from the midline of the display and approach the target objects. Gaze points were averaged into 1 s bins and were projected to the x axis of the display (Figure 3). The averaged eye movements showed a greater proximity to the target object only in the ostensive condition. Thus, we may conclude that dogs' context-dependent responsiveness to human head turning mirrors the specific effect of human ostensive communication on dogs' cognitive processing. Although the stimuli in the two conditions were equally successful in orienting dogs' attention toward the actor's head in the addressing phase, only the ostensive cues led to gaze following. However, in order to elicit a comparable saliency of the addressing phases in the two conditions in the NO condition, we displayed a moving attention-getter on the model's forehead (Figure 1C) similar to the infant
Recent evidence suggests that preverbal infants' gaze following can be triggered only if an actor... more Recent evidence suggests that preverbal infants' gaze following can be triggered only if an actor's head turn is preceded by the expression of communicative intent [1]. Such connectedness between ostensive and referential signals may be uniquely human, enabling infants to effectively respond to referential communication directed to them. In the light of increasing evidence of dogs' social communicative skills [2], an intriguing question is whether dogs' responsiveness to human directional gestures [3] is associated with the situational context in an infant-like manner. Borrowing a method used in infant studies [1], dogs watched video presentations of a human actor turning toward one of two objects, and their eye-gaze patterns were recorded with an eye tracker. Results show a higher tendency of gaze following in dogs when the human's head turning was preceded by the expression of communicative intent (direct gaze, addressing). This is the first evidence to show that (1) eye-tracking techniques can be used for studying dogs' social skills and (2) the exploitation of human gaze cues depends on the communicatively relevant pattern of ostensive and referential signals in dogs. Our findings give further support to the existence of a functionally infant-analog social competence in this species. Results Although recent research has provided important evidence about dogs' social communicative skills [2], it is still unclear whether dogs' gaze following is tuned to cues that signal the human's communicative intent (e.g., eye contact, verbal addressing). In order to investigate this first, we have collected eye-gaze data from 16 adult, task-naive pet dogs. Subjects were presented with a series of movies in which a human female turned her attention toward one of two identical containers either in an ostensive-communicative (O) or in a nonos-tensive (NO) manner. In the ostensive condition, the human actor overtly expressed her communicative intent, whereas in the nonostensive condition, we removed the ostensive signal from the stimulus. Each trial consisted of introductory, addressing, and cueing phases (Figures 1A–1C). All dogs participated in both O and NO trials that were different only in the addressing phase. We obtained valid data for analysis from 13 dogs in the ostensive and 14 dogs in the nonostensive condition; however, only 11 of them provided valid data in each of these conditions (for the validity criteria see Data Analysis). In the addressing phase, dogs spent similar amounts of time gazing toward the human actor in the two conditions (mean 6 SEM: 1,088.8 6 181.1 ms in O and 980.9 6 267.8 ms in NO conditions , ns) and invested a comparable amount of time scanning the region containing the actor's face relative to the whole body: mean 6 SEM = 0.46 6 0.09 in the O condition and 0.55 6 0.10 in the NO condition [paired t test t(10) = 20.88; p = 0.39] showing that in the addressing phase, the human actor evoked the same level of visual attention in both conditions. Next we analyzed whether dogs looked longer at the gaze-congruent area (Figure 1D) as compared to the gaze-incon-gruent area (cumulative accuracy). In accordance with infant eye-tracking studies (e.g., [1]), difference scores were calculated for this variable. We found that subjects looked longer to the gaze-congruent area than to the gaze-incongruent area after having seen ostensive addressing (one sample t test, t(12) = 2.382; p = 0.034). However, this was not the case for the nonostensive condition in which the difference score did not differ from zero [one-sample t test, t(13) = 20.756; p = 0.46], indicating no tendency to follow the human's gaze in the absence of communicative addressing (Figure 2). This differential sensitivity to human referential gestures is strikingly similar to that found in a study of 6.5-month-old human infants [1]. A similar analysis on dogs' first look did not reveal any significant bias toward the gaze-congruent area in O or in the NO conditions [one-sample t test, t(12) = 1.167; p = 0.266; t(13) = 20.105; p = 0.91, Figure 2]. A within subject analysis of the difference scores for cumulative looking time in the two experimental conditions was run on the 11 subjects that gave valid data in both conditions. This analysis shows that dogs were more likely to follow the model's gaze in a gaze-congruent manner in the O than in the NO condition [t(10) = 2.49; p = 0.03; the effect was independent of presentation order; see Supplemental Results available online]. However, no difference was found between conditions for the first look [t(10) = 21.21; p = 0.25]. We also explored the spatiotemporal pattern of eye movements during cueing phase, investigating how the gaze points move away from the midline of the display and approach the target objects. Gaze points were averaged into 1 s bins and were projected to the x axis of the display (Figure 3). The averaged eye movements showed a greater proximity to the target object only in the ostensive condition. Thus, we may conclude that dogs' context-dependent responsiveness to human head turning mirrors the specific effect of human ostensive communication on dogs' cognitive processing. Although the stimuli in the two conditions were equally successful in orienting dogs' attention toward the actor's head in the addressing phase, only the ostensive cues led to gaze following. However, in order to elicit a comparable saliency of the addressing phases in the two conditions in the NO condition, we displayed a moving attention-getter on the model's forehead (Figure 1C) similar to the infant
Uploads
Papers by Kiana Firouz