by Kelly Yee Lai Ku, Catherine Marsh, Karen Webster, Laysha Ostrow, Dana Elfenbein
Sponsors: Dr. Nicholas S. Thompson and David Stevens
Abstract
The literature is full of paper purporting to demonstrate which features or infant cries govern caregiver responses. However, this research focuses on those features of infant cries that are consistent with researchers' methodological convenience or theoretical preconception. In this study, we employ a methodology that permits participants themselves to nominate salient features of cries. Our results confirm traditional observations that pitch, urgency and level of distress are salient features of infants' cries as well as vocal discontinuities such as coughs, hiccups and partially voiceless vocalization. This outcome supports the Respiratory Drama's account of the form of infants' cries.
Introduction
Evolutionary psychologists seek to explain human behavior in terms of its evolutionary function, that is, the selective pressures that have brought the behavior to its present form. In the history of the human species, crying could have been selected only if it had an audience and if that audience responded in some way that benefited the crier, or at least, the genetic substrates of crying. If the design of infant cries was to provoke caregiving responses, then selection should have shaped the cries to efficiently mediate those responses. In return, given a common interest between caregiver and infant, selection should have favored caregiver sensitivity to the properties of the cries. Thus, while infants' cries were being shaped by evolution for their effectiveness in mediating caregiving responses, caregivers were being selected to be responsive to infant cries. Two essentially different ways of thinking about this evolutionary interplay between infants' cries and caregivers' responses have characterized the literature. The first has focused on the information value of the cries. On this account, a cry functions to provide caregivers with information about the nature of the infant's problem -- hunger, fatigue or pain, for instance. This idea has proved to be highly heuristic and has provoked decades of research on infants' cries and caregivers responses, beginning with the classic work of Wazc-Hockert in the 60's. However the recent implication of this research is to suggest that the information perspective is wrong, at least as if it requires that specific features convey specific information to caregivers concerning the infant's situation. An alternative account (Thompson, 1998; Krebs and Dawkins, 1978) is that crying functions not for its information content but for its direct effect upon the nervous system of caregivers. The more distressed the infant, the more its cry takes on properties that are directly disturbing to caregivers and draws their attention to the situation of the infant. This "direct effect" perspective requires ancillary hypotheses to explain why, over evolutionary history, infants' cries have taken on the particular properties they have and why these cries have the effects that they do upon the nervous systems of adults.
Our laboratory has proposed the Respiratory Drama Hypothesis to meet the ancillary requirement. According to the Respiratory Drama Hypothesis the evolution of human infant cries arises from the special vulnerabilities of the human respiratory system imposed by speech. These vulnerabilities led to selection on human caregivers for sensitivity to signs of distress in an infant's breathing and this sensitivity, in turn, led to selection on human infants to project, or even simulate, features of respiratory distress in their crying (Dessureau et al.). On this account, some of the features of infant crying that most capture the attention of caregivers should be related to signs of respiratory distress.
The literature is full of papers purporting to demonstrate which features of infant cries govern caregiver responses (Furlow, 1997; Gustafson, Wood & Green, 2000; Boukydis & Burgess, 1982). But each of these studies has been designed to test the efficacy of properties chosen for methodological or theoretical reasons. We decided to take a step back from the concerns of previous researchers and ask our participants to tell us which properties were salient to them. The method we used was Multidimensional Scaling (MDS), which provides for the identification of salient criteria by modeling similarity judgments of elements in an array of stimuli. With respect to infants' cries, we hypothesize (1) Participants will use a small number of specific criteria in their judgments of similarity amongst cries; that (2) participants' ratings of the qualities of the cries will relate to these criteria; that (3) both ratings and similarity criteria will related to the measured physical variables of the cries; that (4) at least some of the effective properties of the cries will relate to respiration.
Method
Participants
32 Clark University undergraduate students (10 males and 22 females) participated in this study. Ages ranged from 18 to 28 years old. Volunteers were recruited from psychology classes but told that participating would not affect their academic standing in their classes, and that they could quit the experiment whenever they desired.
Stimuli
The recording was made in a home setting where a 3 1/2 month infant male was put down for a nap after a meal. A 15-minute single cry bout from the infant was divided into 16 five-second cry segments, using Sound Designer II on a Mac computer. These cries were presented to the participant sample in all possible pair wise combinations to eliminate the possibility of order effects.
Procedure
The experiment took place in 2 phases approximately a month apart. Both phases were held in a room equipped with a Pentium III computer with Lab Tec speakers. Participants were put through the procedure one-by-one. Each Participant was greeted by the experimenter and was asked to sit in a chair in front of the experimenter and out of sight of the computer screen.
In the first phase, the researcher played the 16 cry stimuli to the participants in order to familiarize him or her with the cry sounds. Participants then listened to 30 pairs of cry stimuli that were randomly generated in advance. They were instructed to rate the similarity of each pair on a scale of 1-100 (1=not similar, 100=identical). A sample scale was taped on the table in front of the participants. Participants verbally reported their similarity ratings. After participants finished rating their pairs, they were asked to write down the answer to the question, "what characteristics of the cry did you use to make your similarity judgments?" on a piece of paper.
In the second phase Experimenters played the same 16 cry stimuli in a computer-generated-random-order through headphones to re-familiarize the participants with the cry sounds. Continuous likert scales were given to the participants. Participants required approximately 3 minutes to rate each stimulus. Stimuli were played at least twice, and more as needed.
Measurement of Physical Properties of Cries
Six of the stimuli that had the greatest distance on the MDS (Multidimensional Scaling) solution (see below and Figure 1) were chosen for the measurement of physical properties. 15 physical properties such as duration, phonation, formants, partially voiceless vocalization, fundamental frequency etc. were measured to uncover the physical basis for subjects' similarity judgments and quality ratings of the cry stimuli.
The physical properties of the cry stimuli were measured by sound analysis programs Canary 1.2 and/or Pratt 3.6.68 running on a G4 Macintosh and/or Pentium III Computer. Each element of each cry stimulus was identified as type (wail, inhalation, exhalation, hiccup or cough) and each unit was further segmented by phonation types (phonated, partially phonated and disphonated segment). These data were entered into a primary database. From the primary database, the means of each property in each sound type are calculated for regression analysis.
Statistical Analysis
Participants' similarity judgments were analyzed by MINISSA, a MDS (Multidimensional Scaling) program. MDS represents participants' judgment of the similarity of cries by transforming the perceived relatedness among stimuli into a visual representation of distance in a spatial model. The number of spatial dimensions and organization - distance of the cries in the perceptual model reveal the structure (features and criteria) on which the judgments were based.
In order to interpret the model, we correlated the co-ordinates of the stimuli in the model with participants' quality ratings and with the means of physical measurements of the cries, using Multiple Regression analysis.
Results
These data indicated strong correlations between participants' similarity judgments, and some perceived qualities and physical properties of the cries.
Dimensions of Similarity Judgments
A three-dimensional solution was chosen, stress = 0.1333, because it produced a satisfactory representation of the judgments (Figure 1.). The three-dimensional solution fits well with the similarity data, because the decline in stress from the two dimensional solution to three-dimensional solution is substantial (.22 - .13 = .09), whereas the decline in stress from three to four dimensions is comparably small (.13 - .08 = .05); thus, adding another dimension on the three-dimensional solution would have little effect in the patterns of the similarity judgments, but increase the risk of over-fitting.
Correlation of Similarity Judgments and Perceived Qualities
The mean ratings from all the 12 qualities were used as predicted (dependent) variables in Multiple-Regression with the MDS 3 co-ordinates as predicting (independent) variables. Amongthe 12, those qualities that showed a regression of r=>.8 are shown in Table 1, below:
Figures 2a & b present the solution as two 2-dimensional arrays with the relevant regression lines from those qualities found in Table 1. Figure 2a shows 3 projected regression lines of qualities: A) Pitch B) Quiet, Energy, Distress & Urgency & C) Perceived Age, on dimension 1 & 2. Figure 2b shows the projected regression line of the quality Coughing on dimension 3. These results suggest the similarity judgments were based on group of qualities defined by ratings of Quiet, Energy, Distress & Urgency that are closely related to quality of Pitch, and the relatively unrelated qualities Perceived Age and Cough. The high multiple Rs (=>.80) indicate that the dependent variables (perceived qualities) predict the pattern of the MDS solution, thus suggest that these qualities underlined criteria determining similarity judgments.
Correlations of Similarity Judgments, Perceived Qualities and Physical Properties
Relationship between perceived similarities and qualities underlying those quality ratings and some physical properties of the stimuli were demonstrated. Among the properties being analyzed, the multiple-regression between the coordinates and percentage of Partially Voiceless Vocalization are strong r =.99, p<0.009. No further significant relationship were found. Partially Voiceless Vocalization predicts the co-ordinates of the three-dimensional solution almost perfectly, hence the three-dimensional space model can be defined in part by the physical properties of Partially Voiceless Vocalization of the cries that were interpreted in terms of Distress and Urgency by the participants.
Further, two significant relationships between quality ratings and physical properties of the stimuli were found: 1) Percentage of Partially Voiceless Vocalization and Distress & Urgency 2) Percentage of Cough and Hiccups (Rs >=0.89, p<0.02). Therefore, discontinuity identified as Hiccupping and Coughing account for a dimension in the co-ordinates of the MDS solution.
Participants' quality ratings and physical properties of stimuli were related with each other. The patterns of the three-dimensional solution also correlate well with the quality ratings, hence, not only did the perceived qualities reflected the physical properties of cry, but the participants used the perceived qualities as the specific criteria for comparing cries.
Discussion
The strong correlations between participants' similarity judgments, perceived qualities and the actual physical properties of cries provide some support for all four of our hypotheses. We conclude that participants used certain criteria to judge how similar are the cries, and these are reflections of the perceived qualities of the cries. And both the similarity judgment and perceived qualities are based on the actual properties of cries. Participants in this study demonstrated the ability to recognize the salient features of infant cries. Those recognized features are the structures underlying the criteria used by participants, which are well represented in a 3-dimensional space model.
We believe that this study has taken an important step in answering the question of what exactly are adults listening while an infant cries. The significance of perceived cry qualities from our participants such as Distress, Urgency, Age and Pitch were also known from previous research. However, in this study these properties are identified by participants rather than being built into the research protocol for technological or theoretical reasons.In addiction, our participants identified 3 new properties: Coughing, Hiccupping and Percentage of Partially Voiceless Vocalization. Interestingly, all three of the new variables, and their relationships to Distress and Urgency seem to related well with the Respiration Drama Hypothesis that suggests the infants were simulating a respiratory distress to manipulate parents' behaviors. Participants in this study paid special attentions to these properties when comparing cries, and were rated cries with these properties as more Distress and Urgent. Future studies should focus on revealing the importance of the three new properties of cry, its relationship to the Respiration Drama Hypothesis and other theories in literature.
References
Boukydis, C., and Burgess, R. (1982). Adult physiological responses to infant cries: Effects of temperament of infant, parental status, and gender. Child Development 53: 1291-1298.
Furlow, F. (1997). Human neonatal cry quality as an honest signal of fitness. Evolution and Human Behavior 18: 175-193.
Gustafson, G., Wood, R., and Green, J., (2000). Can we hear the causes of infants' crying? In Crying as a sign, a signal, and a symptom, Barr, R., Hopkins, B., and Green, J. (Eds.). London: Mac Keith Press, 2000, pp 8-22.
Krebs, J. R., and Davies, N.B. (Eds.). (1978). Behavioural ethology: An evolutionary approach. Oxford, United Kingdom: Blackwell.
Thompson, N. S. (1998). Infant Cries as Evolutionary melodrama: Extortion or deception? Evolution of Communication 2(1): 25-43.
Wasz-Höckert, O., Michelsson, K., and Lind, J. (1985). Twenty-five years of Scandinavian cry research. In Infant crying, B. Lester and C. Boukydis (Eds.). New York: Plenum Press, 1985, pp. 83-104.
Acknowledgement
We are grateful to David Stevens and Rachel Falcon for their help with statistical analyses, and to the National Science Foundation for an Institutional Instrumentation grant to support undergraduate research at Clark.