While the dogs' success in both experiments did not differ between conditions, our detailed behavioral analysis revealed that, when searching in the dark, dogs spent a longer time actively searching and sniffed more.
These findings indicate that dogs integrated information perceived through different sensory modalities and that, while vision was among the preferred modality for identifying the objects tested in this experiment, dogs can spontaneously and successfully revert to using only other senses if visual information is not available. By doing so, dogs present a flexible use of different sensory modalities (see also Szetei et al. 2003; Polgár et al. 2015).
The occasional straight approaches observed only in the light baseline suggest that, when visual information is available, dogs can also identify the object from a distance. However, most often, dogs tended to search among the different objects from a closer distance. This indicates the use of close-range vision and also, potentially, other sensory modalities, including not only olfaction but also touch—as we found very few and short occurrences of sniffing in the light baseline. Our results are consistent with the findings of Bräuer and Belger (2018), who observed that sniffing behavior increased the latency of approaching and decreased the number of direct approaches towards a target object.
Humans can rely on tactile information when visual input is limited (Lacey et al. 2007). Nevertheless, our results did not reveal differences in the time spent by the dogs exploring the toys with their mouth (i.e., mouthing behavior) between conditions in both experiments. This may indicate that these senses are equally used, irrespectively of the illumination, or that they are not relied on at all in object search. However, dogs may also display behaviors other than what we defined as “mouthing” when using tactile or gustatory senses, such as using their noses or whiskers. Thus, we do not exclude that these sensory modalities may have been used differently in the two experimental conditions by the dogs. In addition, dogs often mouth toys as part of their play behavior. It could, therefore, be that the definition of this behavioral variable was not sensitive enough to reflect the use of tactile sensation.
In Experiment 1, all dogs displayed a high success rate, that did not differ between conditions. This demonstrates that both T and GWL dogs can discriminate between a target object, associated with a reward during the immediately preceding training, and distractor objects. These findings are in agreement with previous studies reporting on dogs’ ability to perform object discrimination tasks (Milgram et al. 1994; Head et al. 1998; Tapp et al. 2004) and expand those to situations of limited sensory information. Our finding that although the dogs’ success rate in Experiment 1 was already above chance when tested on the first toy (i.e., toy 1), their performance increased when tested again (i.e., on toy 2), could be attributed to the dogs becoming experienced in the task and familiar with the test situation during the experiments (Hunter and Kamil 1971). Similarly, Bräuer and Belger (2018), described that dogs’ latency of finding a target object decreased as their experience in the task increased.
We did not find differences between the success rate of T and GWL dogs in the object discrimination task, nor did we observe differences in their searching behavior. This suggests that the extreme difference between the ability of GWL and T dogs to recognize objects based on their labels (Fugazza et al. 2021a, b) does not result from differences in object discrimination capacities.
While in Experiment 1, the two groups of dogs discriminated rewarded from non-rewarded objects, in Experiment 2, the objects from which the GWL dogs had to select were all familiar objects. Thus, this is a specific complex case of object recognition that cannot be solved by simply relying on familiarity. The GWL dogs' success in recognizing these objects according to their verbal labels did not differ between dark and light conditions. Ganea (2005) described how, after hearing the names of familiar objects, 14-month-old infants started to search for them and found the objects, thereby demonstrating that the objects’ verbal labels led to the retrieval of the object’s representation. When tested in the object recognition task, GWL dogs demonstrated that they can recognize familiar objects under limited sensory inputs, thereby demonstrating that they have formed a multisensory mental representation of the object (Lacey and Sathian 2011, for review). Moreover, the GWL dogs’ success in retrieving the named toys shows that for each object verbal label, they form a specific multisensory mental representation, enabling them to recognize the correct toy even when it is placed among other labeled objects in the dark. In other words, for GWL dogs, hearing an object’s verbal label evokes a mental representation of the object.
To summarize, we found that, in the absence of formal training, dogs mostly rely on proximate vision and, potentially, touch sense in object discrimination and recognition tasks but can switch to using only other sensory modalities when vision is not possible. Dogs spontaneously encode different features of the objects, leading to the construction of multisensory mental representations. In the case of GWL dogs, a memory of the multisensory representation is evoked by hearing the objects' verbal labels as they perform complex object recognition tasks.