*5.4. Results*

Figure 7 shows the differences in experimental performance according to the degree of visual impairment. In this analysis, a Tukey Honestly Significant Difference (HSD) test was conducted to examine the significance of the differences (*p* < 0.05). In the case of Type *α*, both the identification rate and recognition rate were significantly different for the visually impaired compared to the sighted or blindfolded sighted. In other words, the null hypothesis that there is no difference in the ability to recognize sounds between sighted and visually impaired was rejected. This shows that the people with visual impairments had a better ability to discriminate and localize sounds in verbal and representational sounds than sighted people and blindfolded people. In the case of Type *β*, there was no significant difference in either the identification rate or the recognition rate.

**Figure 7.** Differences in performance by degree of visual impairment.

Figure 8 shows differences in experimental performance according to the audio type. In this analysis, a Tukey HSD test was conducted to examine the significance of the differences (*p* < 0.05). In the case of Type *α*, the abstract sounds were significantly different from the other two audio types for all participants in both the identification rate and the recognition rate. In other words, the null hypothesis that the effect of the abstract sound on the identification/recognition rate of the player is not different from that of the other two types of sound was rejected. This suggests that verbal and representational sounds were effective in this system in presenting the content and location of the audible cards for all experimental participants. In the case of Type *β*, there was no significant difference in either identification rate or recognition rate. As a general tendency, representational sounds tended to obtain higher identification and recognition rates. In addition, we obtained the opinion from the participants that "Verbal sounds were mixed up and became unintelligible when they were emitted at the same time". These results sugges<sup>t</sup> that it may be inappropriate to use verbal sounds in Type *β*.

**Figure 8.** Differences in performance by audio types.

Figure 9 shows the differences in experimental performance according to the style of sound playback. In this analysis, Welch's *t*-test was conducted to examine the significance of the differences (*p* < 0.05). Type *β* has the advantage that players can obtain the information of multiple audible card at once compared with Type *α*. There was a significant difference for all participants in both the identification rate and the recognition rate compared to Type *α*. In other words, the null hypothesis that there is no difference in the effect of the style of sound playback on the player's identification/recognition rate was rejected. This suggests that Type *β* is not effective for accurately recognizing multiple audible cards.

**Figure 9.** Differences in performance by the style of sound playback.

### *5.5. General Discussion*

This experiment was conducted to clarify the effect of each valuable parameter of the proposed system on the player's game performance. We prepared parameters for three types of degree of visual impairment, three audio types, and two types of the style of sound playback.

Comparing the experimental results according to the difference in the degree of visual impairment showed that the people with visual impairments had a better sound discrimination and localization ability than the sighted people. In addition, a better sound discrimination and localization ability of people with visual impairments has been mentioned in previous studies [43,44]. Therefore, people with visual impairments were able to recognize the audible cards as well as or better than the sighted people. In other words, we found that sound presentation was effective as a solution for presenting public information indicated by components in board games.

Comparing the experimental results according to audio types, the results showed that representational and verbal sounds enabled players to accurately recognize the information indicated by the components. Although the verbal sounds were able to convey the information correctly to all the participants in the experiment, one participant commented that the verbal sounds prompt confusion when they are produced simultaneously. In addition, the fact that verbal sounds are easily masked by background noise has been mentioned in previous studies [45]. Thus, it is undesirable to use multiple verbal auditory stimuli in this system. Further, verbal sounds should not be used in board games where verbal communication is active, because players can miss the information indicated by the components. In other words, verbal sounds should only be used as audio feedback at times when verbal communication is not occurring, such as before and after a competition. As with the verbal sounds, the representational sounds were able to convey information correctly to all participants. This shows that representational sounds have a grea<sup>t</sup> potential to be used effectively in an auditory card game system. In addition, they are superior in terms of intuitiveness, learnability, memorability, and user preference according to previous work [45–47]. These advantages sugges<sup>t</sup> that representational sound has the potential to make board games even more fun. Compared to verbal sounds, they are less likely to be masked by background speech [45], so they are more effective when presenting multiple sounds stimuli simultaneously. When representational sounds cannot fully express information, abstract sounds should be used. Since abstract sounds are not easily masked by background sounds as well as representational sounds [45], they can also be effective when multiple auditory stimuli are required. However, unlike representational sounds, abstract sounds are obviously limited in the size they can be learned. The experimental results obtained in this study also confirmed that abstract sounds can cause recognition errors. According to the experimental results, when abstract sounds were used as stimuli, the maximum number of audible cards correctly recognized by the participants was from about 5 to 8, regardless of the degree of visual impairment. Similarly, previous studies [41,48] have recommended limiting the set of abstract sounds to a maximum size of 5 to 8. For these reason, it is appropriate to keep the number of abstract sounds within the range of 5 to 8. We recommend the use of abstract sounds as feedback for player actions and changes in the game, rather than indicating component contents.

Comparing the experimental results according to the style of sound playback showed that using many cards at the same time has a negative impact on recognizing components. The number of audible cards that the participants could recognize correctly was about 3, even if the auditory stimuli were given at the same time. Therefore, it was suggested that the players could recognize the audible cards correctly if the number of these was about 3. However, verbal sounds are easily masked by background speeches, so representational/abstract sounds should be used when multiple audible cards are presented simultaneously.

Finally, in order to clarify what caused the errors of the participants, we investigated the error composition in the experiment. Errors consist of the following three types.

1. Sound identification error: experimental participants misinterpreted the meaning of the audible card presented them.


Figure 10 shows the average number of errors per trial for the experimental participants. In this analysis, a Tukey HSD test was used to examine the significance differences (*p* < 0.05) in the number of errors between players. The null hypothesis is that there is no difference in the average number of errors between players.

**Figure 10.** The average number of errors per trial for the experimental participants. There were three types of errors: sound identification errors, vertical localization errors, and horizontal localization errors.

Sound identification errors were generally not significantly different regardless of the audio type. However, there was a significant difference between the visually impaired and the blindfolded in the representational sound. This suggests that the presence/absence of visual information can have an effect on the sound discrimination itself.

The number of vertical localization errors for the visually impaired was significantly lower in verbal and representational sounds than for the sighted. The reason for this result is considered to be that the speaker of the iPod touch is located at the bottom of the device. It was predicted that the sighted people were influenced by the visual information and mistook each device for the lower device. However, the number of errors for the visually impaired was significantly lower than that for the blindfolded participants, who were excluded from visual information. In other words, people with visual impairments showed a higher vertical directional resolution than the sighted people. Ohuchi [43] and Voss [44] concluded that people with visual impairments tend to have higher horizontal directional resolution and distance resolution than sighted people. Therefore, the results of the high vertical resolution of the visually impaired in this experiment can be attributed to their high distance resolution. However, in the vertical localization of abstract sounds, people with visual impairments did not show a significant difference in the number of errors compared to the sighted people. Therefore, the results sugges<sup>t</sup> that people with visual impairments can have difficulty with the source vertical localization of abstract sounds compared to verbal and representational sounds. In other words, the use of abstract sounds in the proposed system can interfere with the high vertical resolution of people with visual impairments. As a result, it is not desirable to use abstract sounds in situations where we want to ensure the vertical directional resolution of visually impaired people, such as when we want to accurately convey the position of a component. However, abstract sounds have

the advantage of being freely designable, so reflecting the component's position in the rhythm, pitch, and timbre of the sound could make it easier to locate the component.

Horizontal localization errors did not differ significantly among participants, regardless of the audio type. Horizontal localization seems to be easier for all subjects because it reduces misrecognition of sighted participants due to the position of the speaker and increases auditory cues such as interaural level difference and interaural time difference. Therefore, if the component's position is to be shared by all players, it is recommended that the components be aligned as horizontally as possible. Furthermore, it is possible to control the placement of the cards to give an advantage to people with visual impairments.

To allow for more flexible board game design, it is also possible to include visual information on the audible cards. At this time, it is easy to assume that visual stimuli affect the recognition of components to sighted people. In fact, in previous studies [49], it has been concluded that the type of sound and its congruency with visual information can affect reaction time. Therefore, it is possible to induce confusion by providing visual content that differs from auditory content to sighted people.
