Let’s now consider in more detail the deficits in vocal communication associated with ASD. Social interactions critically involve individuals to both vocally express their emotions and to understand and respond to the emotional expressions of others. The emotions experienced by an individual can be expressed through the semantic content of what is said. If I tell you that I feel angry, despondent, frustrated, defeated, overwhelmed, cagy, indignant, exhilarated, content, proud, or surprised, you have an idea of how I feel. Emotions are also critically conveyed by how words are said; that is, through vocal prosody. Prosody can be described at multiple levels. At the acoustic level, it refers to the usage of features, such as variations in the pitch of a phrase or sentence, the duration or loudness of a particular syllable, or vocal quality (e.g., tense, hoarse, breathy). At the mechanical–physiological level, prosody refers to how these acoustic features result from the complex interplay of all components of the speech production apparatus, ranging from the lips to the lungs. At the brain level, prosody refers to how these components, in turn, are controlled in a coordinated fashion by multiple brain systems. Finally, at the functional level, we can delineate the following classes of functions. Affective prosody refers to the expression of internal emotional experiences of joy, sadness, achievement, or anger. For example, a flat pitch contour can convey sadness, while significant dynamic changes in pitch and energy levels can indicate greater emotional salience. A tremble in a voice can indicate sadness or fear.

Grammatical prosody helps us to distinguish between meanings of words or sentences. Stress on a particular syllable can help us to differentiate a verb (pre-sent’) from a noun (pre’-sent) and the different meanings of a word, while upward trend in pitch can help us identify a question from a string of statements. Pragmatic prosody is used to help us put emphasis on an important idea, sentence, or word in a sentence and thus reveals our focus and intentions. Enhanced duration and loudness (energy) can indicate the salient point of a statement. For example, “It wasn’t the blue chair you spilled ink on, it was the NEW chair.” Prosody allows us to express subtle or not so subtle attitudes. For instance, one can be sincere or sarcastic depending on how one says, “I really want to go to dinner with you.” In another, and quite different function of pragmatic prosody, people speak to infants in “motherese” or “infant-directed talk” (11) which they do not use with older children. People speak differently to a policeman than they do to an intimate partner. Pragmatic prosody thus also signals the context, roles, and relationships between speakers. Although it is customary in human research to include role-related prosody in the category of pragmatic prosody, we will include it in the category of affective prosody for studies of rodent vocalizations. While role-related prosody is relevant for mice, its classification as “pragmatic prosody” would be inappropriate because of its strong verbal connotations and because, even in humans, role-related prosody typically has an affective component (e.g., motherese cannot be de-coupled from the feelings toward the baby).

Atypical prosodic expression has been considered a hallmark of autism since the disorder was first described (12). Prosody in ASD has been described as monotonous or flat (13), stilted, robotic, overly melodious, or singsong (14). Atypical features include aberrant stress, unusual pitch patterns, abnormalities of rate and volume, and problematic quality of voice (e.g., hypernasality) (15, 16). Children with ASD are more likely to express odd phrasing patterns and to exhibit abnormalities in their modulation of pitch, and in their rate and volume of speech compared to typically developing children (17). Unusual prosody in very young children, ages 12–24 months, has been found to be one of the “red flags” that distinguish ASD from other developmental disabilities (18).

Unusual prosody may interfere with the communicative effectiveness of speech, for example, by making it difficult to determine whether or not a statement is intended as a question, or what emotion is being conveyed. In fact, certain repetitive patterns of specific pitch, duration, or energy changes irrespective of verbal content may trigger an unintentional negative emotional response in the listener, which may then severely disrupt the reciprocity chain. In ASD, atypical prosody very much impacts communicative competence and reciprocal social interactions.

Using objective methods, expressive prosody issues have been documented in about 50% of a group of adolescents and young adults with ASD (16). Receptive prosody issues have been less frequently studied, but are thought to be closely associated with expressive deficits, both having been looked at in autism in studies that use controlled perceptual methods (17, 18). Automated methods for the study of expressive prosody are few and will likely expand on the labor-intensive approaches pioneered in seminal studies, such as Shriberg’s PVSP (16, 19). Manual approaches unavoidably reduce the amount of data available for research (e.g., it is not unusual to analyze only the first 50–100 utterances of an individual’s language sample, even when there may be many more relevant utterances in a spontaneous conversational dialogue across different topics or activities, for example, using the ADOS). Second, the use of the human ear in the PVSP and in Peppé’s methods introduces additional variables, particularly for detection of more subtle cases of atypical prosody (9). Clinical judgment ratings on the ADOS for atypical prosody, for example, have such poor inter-rater reliability, that this item, despite its importance, has never been included in any version of the ADOS “algorithm,” that subset of items most critical for a diagnosis.

Advances in speech technology, mathematical analyses, and algorithm development, however, are making it increasingly possible to perform automated analyses of expressive prosody. For example, preliminary findings using automated analysis of expressive prosody show that even when in ASD certain prosodic contrasts (e.g., focal stress) are expressed clearly; they are expressed with an atypical usage of prosodic features that are not detected by human observers (9). Speech technology can also create stimuli for the study of receptive prosody with carefully controlled or manipulated speech stimuli that cannot be produced by humans, such as affectively distinctive speech stimuli that have identical speaking rates, or stimuli in which certain prosodic features (e.g., pitch) are altered while keeping others (e.g., duration) invariant (20, 21). These speech manipulation methods can in principle be applied to nonhuman vocalization as well, for example, to determine which acoustic features are behaviorally relevant when studying different strains of mice in different situations.

Nonverbal communication is thought to be a critical component of social interaction and communication of affect. Evidence exists that facial and gestural communication are impaired in ASD, both expressively and receptively, as is the use of eye gaze.

ASD has been linked to impairments in the ability to perceive others’ emotions, intentions, and mental states, as well as in the ability to perceive that the experience of another individual is different from one’s own. These concepts are embodied in the notion of empathy. In the early twentieth century, empathy was coined from the German word, einfühlung, or “feeling into” (54). This term has since been expanded and refined by numerous authors (55, 56), but contemporary definitions of empathy continue to maintain a primary role for affective reactivity to others (57). Eisenberg (58) defined empathy as “an affective response that stems from the apprehension or comprehension of another’s emotional state or condition, and which is similar to what the other person is feeling or would be expected to feel” (p. 72).

The mechanisms that underlie deficits in empathic ability in autism are far from well understood and are being researched at both neural and behavioral levels. Some studies focus on impaired and atypical social information processing at the behavioral level and the cascade of events that it can induce over time. These deficits may be tied to (both precursor and consequent) functional and structural brain abnormalities. There is research that relates many of the known social deficits in autism to dysfunction in the “social brain” (which includes, especially, the amygdala, fusiform gyrus, limbic system, and areas of prefrontal cortex) (59, 60). Other studies, as mentioned, are investigating dysfunction in the mirror neuron system and its interconnected circuitry, especially as related to deficits in imitation ability to understand others’ intentions, and empathy (48). There has also been important evidence of uneven distribution of white matter and neural under-connectivity in recent research. For instance, studies have shown functional under-connectivity, measured using correlations between time series of pairs of regions of interest in fMRI (61), as well as structural under-connectivity (62). Uneven distribution and volumes of white matter in ASD may contribute to this reduced interconnectivity (63). White matter volumes may be significantly (25% or more) larger in certain brain areas (e.g., the frontal lobe) and smaller in other areas (e.g., the corpus callosum). The uneven distribution of white matter, together with atypically larger, but not functionally more productive areas, points to inefficient interconnectivity. In addition to these findings, there are additional features at the microscopic level that may also have implications, such as increased numbers of minicolumns (vertical columns through the cortex layer) with greater cell dispersion in autism (e.g., Casanova 64).

Implications of under-connectivity include impairments in the integration of complex social information. Understanding of others’ emotions and the processing of other relevant nonverbal social cues may itself be thwarted by the lack of exposure and focus in autism on the relevant information. For example, Klin et al. (5) employed eye-tracking technology to better understand how individuals with ASD view the world. In a series of experiments in which participants watched video clips from Edward Albee’s Who’s Afraid of Virginia Woolf, individuals with autism were found to focus significantly less time on the eye region of the actors’ faces and significantly more time on the mouth, body, and object regions in the film compared to normal controls. Eye-tracking studies in toddlers with ASD have shown, similarly, a preference to look at the mouth rather than eyes, objects over people, and to orient to physical rather than social contingencies (6). Because the nonsocial focus of visual pursuit takes place at a very young age, children then miss out on much socially relevant information and social learning that would allow for greater understanding of interpersonal contexts and improved social adaptation. However, it is unknown to what degree such biases and lack of input is causal, leading to further neural and neurodevelopmental changes, or themselves caused by fundamental neural impairments in brain circuitry.

Under-connectivity may also result in poor cross-modal integration in ASD, which has been documented via behavioral, anatomic, electrophysiological, and neuroimaging methods. At the behavioral level, Wetherby et al. (18), for example, used the systematic observation of red flags (SORF), rated by two people with good inter-rater reliability (k=0.94) to confirm the presence of six red flags that distinguish children with ASD during the second year of life from those with DD (developmental delay) and TD. One of those flags was the lack of typical “coordination of gesture with eye gaze, facial expression, or vocalization” (with at least three modalities coordinated) and was found in 100% of the children with ASD.

As with the above explanation underlining the lack of vital social input early on, there is strong plausibility for poor cross-modal integration being a major obstacle to empathic ability, since empathy requires a near-automatic integration of many subtle behavioral cues in multiple modalities simultaneously. Although behavioral observations have testified often enough to the lack of linkage in multiple verbal and nonverbal modalities, most of these studies have used subjective judgment or manual coding methods. There are, however, studies underway to examine this in an objective behavioral paradigm and to automate detection of affective communication in individual modalities (i.e., facial, gestural, postural, verbal, and vocal prosody) and analyze, via mathematical algorithms, their temporal coordination and synchrony.

As we transition here to a discussion about how we assess mouse models that may be relevant to autism, we can see the tremendous translational benefit of analysis tools that assess gestures and vocalizations and behavioral responses to these social signals. Analogous diagnostics can be developed, in theory and in practice, for studying laboratory mice. Do mice share these nuanced social abilities? In the subsequent sections, we will explore the social repertoire of the feral and laboratory mouse, including a broad assessment of the expressive and receptive aspects of mouse communication.

When a genetic linkage to autism is found, mouse studies can help to discern whether there is a causal relationship between the mutation and changes in brain function or behavior relevant to autism. Mice are useful species for genetics research because they have a mammalian brain and a relatively short generation time (10 weeks from conception to reproductive maturity). The entire mouse genome has been sequenced, mouse genes can be deleted, gene expression can be changed (65), and human genes can be inserted into the mouse genome (66, 67). With these tools, relationships between genetic mutations and social behavior can be examined. After mutations in fmr1 were linked to fragile-X disorder (68), a knockout mouse strain (fmr1KO) was generated to model the disorder (69). The fmr1KO was then studied for relevant cognitive, social, and seizure-related deficits (70–73). As a result, this mouse model is now being used to explore pharmacological and genetic treatments for the symptoms of fragile X (74–76). Other findings of genetic linkages to autism have resulted in knockout mice with disruptions in WNT2 (77), MECP2 (78), engrailed 2 (79), CAPS2 (80), and neuroligin-4 (81).

Are mice capable of social interactions that would make the study of mouse behavior relevant to autism? Feral mice have been found at population densities of up to 4,000 mice per acre (82, 83). In the wild, both males and females establish territories and social hierarchies (84–86) and compete for food, territory, and mates (87). Dominant males can distinguish mice that belong in their territory and the degree of relatedness of mice in adjacent territories (86). In natural environments, social arrangements are dynamic, fluctuating with season and food availability (88). Aggressive behavior is more common at high population densities, while gregarious social groups form when the resources for each mouse are more plentiful (89, 90). Social arrangements are also dynamic, changing with the birth of new litters, with movements of mice into and out of geographic areas, with the death of dominant males, and changing food and mate availability.

One could imagine that a mouse with an impaired ability to recognize expressions of dominance, submissiveness, or the changing needs of offspring might have difficulties surviving in a natural population. For instance, a mother is behaviorally responsive to the wriggling calls (91) and distress calls (92) of her pups. One could envision that a nonresponsive mother would be less likely to produce viable litters. Inability to recognize some of the cues that dictate social hierarchies might also result in forced expulsion from the territory of a dominant male or female. A mouse that avoids social interaction might remain ignorant of social cues for safe foods to eat, safe times and places to forage, and mate availability. Taken as a whole, the natural environment likely imposes strong selective pressures for abilities to detect fear or pain in another mouse and recognize the conventions of social hierarchies (for an enlightened overview, read Crowcroft 93).

Against this backdrop of mouse social capacities in natural environments, we can consider the most useful and widely employed approach for examining the social capabilities of a laboratory mouse; the social approach test. The social approach test has been variously attributed as measure of mouse “sociability,” “sociality,” “reciprocal sociality,” or “social interaction.” This social interaction test has been used extensively to examine mice with targeted alleles relevant to autism (see Section 2). Social approach behavior of a “test” mouse is measured typically after a period of social isolation. During isolation, the test mouse becomes habituated to a cage where it resides. A stimulus mouse is then introduced to the test mouse’s cage. During the test, the stimulus mouse explores the cage and the test mouse follows or “approaches” the stimulus mouse. “Sociality” in this test is defined as a composite measure of social approach; the fraction of time that the nose of the test mouse is within a short distance of the head, flank, or anogenital region of the stimulus mouse. Social approach is most vigorous after extended periods of social isolation and can vary with the time of day when social isolation occurs (94). This test has various permutations. Among juveniles, it can indicate “social interaction” (95). Among adult males or nursing females confronted with a visitor, this resident–intruder test measures aggressive behaviors (96).

It should be noted here that some very insightful information are gained about the social approach behaviors of a laboratory mouse by simply observing the home cage. For instance, dvl1 knockout mice do not huddle together or barber the whiskers of their cagemates like their wild type controls (97). This particular study highlights the importance of simply observing the home cage, in the colony room during periods when the white lights are on and also under dim red light (see details on mouse husbandry at the end of this chapter).

There are also automated versions of social approach tests, to rapidly determine whether a test mouse is more likely to approach an unfamiliar mouse, versus a familiar mouse, or versus an inanimate object (98). Approach behavior is indicated when the test mouse breaks an infrared beam and enters a room that contains a cage wherein the stimulus mouse resides. A more detailed description of this test is provided at the end of the chapter.

Just as social behaviors and motivations of children are different from those of adults, mouse behaviors likewise change with development. For instance, social approach behaviors of early adolescent mice can be very sensitive to genetic background and insensitive to the gender of the test mouse or the stimulus mouse. As mice approach reproductive maturity, social approach becomes more stereotypical: males approach females more than females approach males. Genetic influences of social approach behaviors may be marked by gender differences that emerge with reproductive maturity (99). There are several excellent references on the ontogeny of mouse behavior that should be considered (100–104). A common mistake is to consider a juvenile mouse as a smaller version of a juvenile rat, which is a very different rodent (105–108). For instance, social play among young mice is very infrequent and not at all robust (106), whereas young rats express “rough and tumble” play (108, 109). Further, mouse motor behaviors do not indicate strong “reciprocal” social interactions in this test. The stimulus mouse typically walks around the perimeter of the cage while the test mouse follows the stimulus mouse. These behaviors do not indicate play or reciprocity. It is possible that there are reciprocal vocal exchanges, but this aspect of their social exchange is currently unknown.

It can also be difficult to interpret differences in social approach response. Diminished approach toward another mouse could indicate shyness, social anxiety, reduced preference to engage in a social encounter, or an increased preference for social isolation. Shyness and social anxiety are not core features of autism. Preference for social isolation may have more to do with autism, at least as originally described and emphasized by Kanner (110) who characterized “autistics” as “aloof,” “indifferent,” and “isolated.” Currently, however, the spectrum of autism disorders is much broader and involves a more heterogeneous group of children for whom aloofness or lack of social interest may not be descriptive. Reduced social motivation is not mentioned as a core symptom of the DSM-IV, but it does characterize the broader autism phenotype as described by Dawson et al. (111).

Moreover, in humans “impaired social interaction” has a much more inter-subjective connotation than does diminished social approach, sociability, or “sociality.” A person with autism can be sociable and still have a severe impairment in social interaction. Social interaction has to do with the behavior of one individual impacting another individual, of individuals in interaction. At the very least, impaired social interaction assumes that social behavior is “inappropriate” in the means, timing, or manner in which it is done with respect to both norms (e.g., age, sex, cultural, contextual, social) and with respect to the individual to or for whom the behavior is directed. A question then arises, does the social approach test paradigm capture core symptoms of “impaired social interaction” in autism? Is it reasonable to think that it could, given some modifications to the paradigm?

Can we be sure that a diminished social approach behavior indicates a lack of social motivation? Approach behaviors of any sort toward a stimulus might suggest that the stimulus induces a psychological expectation of reward or pleasure, but this inference is far from guaranteed. For instance, an amoeba, which lacks a nervous system altogether, swims up a chemical gradient. Having a brain allows an animal to navigate through its environment to gain access to rewards and avoid painful stimuli, through the cues that are associated with these states. Such cues can be contextual (berries to eat or thorns to avoid) or temporal (a sting following a buzzing sound). This capacity to make associations allows an animal to adapt to dynamic conditions. For example, an animal can learn that a novel red berry is not sweet. By contrast, an amoeba cannot modify its behavioral strategy to obtain food. Its movement is dictated by membrane receptors, directed by fixed stimulus–response action patterns.

But how do we infer whether a mouse finds a stimulus rewarding or aversive? The question of how to infer animal motivation from animal behavior is a hard problem but it has been extensively studied. Through conditioning experiments (described below), we can infer animal capacities for fear and fear learning (112–114) and for the anticipation and consumption of rewards (115, 116). These capacities are supported by a highly differentiated limbic anatomy (117) and physiological systems that include glucocorticoids, dopamine, serotonin, and endogenous opiates (118). Drugs of abuse can co-opt these natural reward systems (119–123). Thus, mice respond to their environment, in part, based upon associations between particular contextual or temporal cues and underlying affective states, such as feelings of reward or fear.

How can we explore the relationship between social approach and these underlying experiences of reward or pain? The conditioned place preference (CPP) test is particularly valuable in this regard. CPP tests were developed in the context of a rich theoretical framework (120, 124–127). If a particular stimulus (such as access to morphine, cocaine, or a highly palatable food) is rewarding, then a test animal will choose to spend its time in an environment that it associates with access to this putative reward. In the CPP experiment, the test mouse is exposed to two environments discernable by distinct but unimportant qualities, such as particular beddings or odors. One environment is repeatedly paired with the presence of the putative reward and the other environment is conditioned by the absence of the reward. During the test, the mouse is allowed to roam between both of the conditioned environments. No reward is provided during the test period, but the mouse is expected to spend time in the environment formerly paired with the putative reward and less time in an environment where it either did not experience the reward (for example, saline or standard lab chow) or where it previously experienced something aversive. CPP tests show that exposures to morphine, cocaine, and highly palatable foods can be rewarding for rodents (120). CPP tests have also demonstrated that different kinds of social encounters, including play (126, 127), sexual interactions (128, 129), juvenile social interactions (130), mother–infant bonding (131), and even aggression (132), are rewarding to mice.

Within the context of the social CPP, we can ask whether social approach can be associated with social reward. The experimental framework of a social CPP is depicted in Fig. 7.1. BALB/cJ mice (BALB) express levels of much lower social approach than C57Bl/6 (B6) mice, consistent with previous studies (133, 134). The B6 strain shows a strong place preference for environments paired with social access. By contrast, the BALB strain shows no conditioned respond to the different environments; they do not find access to other juveniles rewarding. Comparison of these two strains indicates that social approach can be associated with social reward (130). A more detailed description of the social CPP test is described at the end of the chapter.

We can also envision situations in which the association between social approach and social reward does not exist. As mentioned earlier, a mouse strain might have a strong motivation for social novelty but an aversion to group housing. A mouse might avoid social approach for other reasons, such as shyness or social anxiety, yet prefer social housing to isolation. Social reward can be multiplicitous and ephemeral. Nursing dams, when they must choose between environments paired with access to newborn offspring versus access to cocaine, choose environments paired with offspring during the first week after birth, but when they have older pups, they prefer environments paired with access to cocaine (131).

Social reward is likely mediated, at least in part, by endogenous opioids, dopamine, and their receptors within the brain. These endogenous molecules are distributed within regions involved in emotional regulation and mediate responses to pain and stress, reward, and social bonding (135, 123). Activation of opioid receptors promotes maternal behavior in mothers and social play among juvenile animals (136–138) and can be modulated by agonists and antagonists of opioid receptors (139, 140). Separation distress, exhibited by archetypal behavior and calls in most mammals and birds, is reduced by opioid agonists and increased by opioid antagonists (141). Access to social encounter can, in turn, influence opioid physiology. For instance, social play and short-term social isolation alter opioid receptor binding in the nucleus accumbens and other regions in rats (142). Dopamine is also important in social interactions. In the nucleus accumbens, dopamine appears to be a necessary signal for pair-bond formation in male prairie voles (143). Social stress or social isolation promotes long-term changes in mesolimbic dopamine (144), and social isolation can render animals more sensitive to dopamine-releasing drugs of abuse (145, 146). There are also important roles for oxytocin and vasopressin in social approach behaviors (147, 148), which influence dopaminergic reward circuits (149). While endogenous opioids and dopamine can moderate social approach and response to social access, the interaction of these physiological mechanisms with social encounter does not rule out the possibility that mice, under some circumstances, might engage in or avoid social approach for other reasons, such as novelty-seeking, shyness, or social anxiety.

How can impairments in social interaction be examined in a way that captures mouse social capacities and has optimal translational value for autism research? Let us first consider the mouse social approach test within the context of the DSM IV-TR diagnosis. In what follows, we review the defining features of “qualitative impairment in social interaction,” as listed in the DSM IV, and ask within this set of criteria whether existing or new behavioral tests can be employed to interrogate behaviors relevant to autism. Let us look at a few of these manifestations and how they could be captured in our mouse models. Importantly, in considering these various tests, it is important to maintain a mouse colony and a behavioral testing environment that minimizes variation in mouse social behavior.