Syntactical information

There are at least three ways that we can begin to understand the "meaning" of the chick-a-dee call from the perspective of syntactical information. The first, pioneered by Hailman et al. (1985), is a structural analysis of syntax per se to determine the hypothetical information that may reside in a signaling system. The second is a natural-history approach, documenting calls produced in natural settings and what behavioral or other contextual correlates exist (e.g. Smith 1972). The third is an experimental analysis, using playbacks to test predicted information content in the calls (e.g. Freeberg and Lucas 2002; Clucas et al. 2004; Templeton et al. 2005).

13.2.5.1 Syntax

Hailman et al. (1985) measured the amount of information that might be conveyed by the chick-a-dee call system of black-capped chickadees. From Shannon and Weaver's (1949) theory of communication, a communicative system with four distinct units (note types) has a maximum uncertainty of two bits of information (UM = log2 N, where N = number of units). The maximum information in a communicative system occurs when all of the system's units are used with equal probability. As is the case with letters in the English language (e.g. relative use of "e" vs. "x"), however, note types in chick-a-dee calls are not used equally—D notes are far more common. This means that the actual information in terms of note use in chick-a-dee calls is lower than the maximum possible information. The uncertainty measured for actual use of the different units of a communication system is referred to as the zero-order uncertainty, and is calculated as:

where Pi is the probability of each of the i units occurring in the system. In the case of the chick-a-dee call of black-capped chickadees, there is indeed a drop-off from UM to U0. However, a much greater drop-off in uncertainty occurs when one assesses the transition probabilities between pairs of notes. This measure of first-order uncertainty, U1, represents the uncertainty of a given unit to occur in a sequence when a previous unit has already occurred. It is calculated as:

where Pi. is the probability of the i and j note occurring in the ijth sequence, and P. ^ is the conditional probability of the j unit occurring given that the i unit has occurred. For the chickadee call this analysis would address the ability to predict that, for example, a D note will follow if a C note occurs in a call. Hailman et al. (1985) found that there is a considerable drop-off in information at this level of uncertainty—if a researcher (and, presumably, a chickadee receiver) detects one note type in a call, there is a good probability of predicting what the next note type will be in the call.

The preceding discussion leads into one of the other major structural features of the chick-a-dee call. Notes and pairs of notes do not occur with equal probability. Instead, the chick-a-dee call obeys rules of note ordering, a simple form of syntax. In black-capped chickadees, the two most common call structures are [A][D] and [B][C][D], with brackets indicating that the particular note type can occur more than once. In other words, if the following notes occurred in a nine-note call, BBCCCDDDD, they would virtually always (e.g. over 99% of the time) occur in the order [B][C][D]. Taken together, the chick-a-dee call represents an interesting case of an open-ended communicative system that is nonetheless constrained by its note ordering rules.

In addition to the constraints upon the call imposed by the note ordering rules, other constraints appear to limit the diversity of potential call structures (Hailman et al. 1987). For example, as the number of A, B, and C notes increase in a call, the number of D notes that might occur decreases. Therefore, there seems to be a constraint on the overall number of notes that can occur in an average call. However, calls with extremely large numbers of D notes are more common than expected by chance, suggesting that the constraints on introductory notes are relaxed when calls contain many D notes (Hailman et al. 1987).

This mathematic approach to the question of information in the chick-a-dee call was extended to another species, Mexican chickadees (P. sclateri; Ficken et al. 1994), and interesting comparative results emerge. The Mexican chickadee chick-a-dee call system is open-ended, the C note is more common than the D note, and the B note is extremely rare. Notes follow the A-B-C-D note-ordering rule shared by black-capped chickadees. The most common call structures were [A][D], [C], and [A], and calls tended to be shorter in note number than black-capped chickadee calls. Ficken et al. (1994, p. 80) indicate that, relative to the chick-a-dee call of black-capped chickadees, the "rarity of B notes and the shorter note length of calls means that the Mexican chickadee's utterances tend to be syntactically simpler, although not necessarily semantically simpler ... " This quote nicely captures the two approaches to information discussed earlier. For information as a mathematically-defined measure, Mexican chickadee calls appear to convey less information than black-capped chickadee calls. Mountain chickadee calls also appear to convey less information than black-capped or Carolina chickadees because their calls are substantially shorter (three-four notes/call vs. six-eight notes/call, respectively; Bloomfield et al. 2004). For information as meaning, however, it is an open question in these species as to which call system conveys more information.

Hailman et al. (1987) published an additional method of evaluating the syntactical properties of black-capped chickadees' calls. They compared the cumulative number of calls containing at least some number of A, B, C, and D notes

(a "survivorship" plot) with the expected cumulative number of notes based on a semi-Markovian model. A departure from the simple semi-Markovian expectation implies some meaning in the note composition of the call beyond a simple repetition of notes. They found that A notes fit expectations almost exactly. B notes did not: calls with three or fewer notes fit expectations whereas those with four or more B notes were too common. As with A notes, the probability of repeating a C note was constant, albeit somewhat less than expected by chance. The survivorship curve for D notes departed completely from semi-Markovian expectations, with too many short-D calls, too few intermediate-D calls, and too many long-D calls. Hailman et al. (1987) suggested that this distribution represented a compound of two or more separate processes, and potentially separate syntactical functions.

We repeated Hailman et al.'s (1987) analysis with a preliminary data set of 2153 Carolina chickadee calls recorded in non-manipulated field settings in eastern Tennessee (nine sites) and central Indiana (six sites). The uncertainty measures for Carolina chickadees for this sample of the field recordings showed a similar pattern to those reported for black-capped chickadees (Hailman et al. 1985) and Mexican chickadees (Ficken et al. 1994), with a marked reduction between zero-order and firstorder uncertainty (UM = 2, U0 = 1.49, U1 = 0.63). A and D notes did not meet semi-Markovian expectations (Fig. 13.3). Long strings of A notes (>six A notes/calls) were more common than expected by chance. The survivorship curve for D notes was qualitatively similar to that of black-capped chickadees: too many calls with a few notes, too few with a large number of notes (10-25), and too many with a very large number of notes.

We asked whether our chick-a-dee calls met expectations of Mandlebrot's modification of Zipf's Law (see Hailman et al. 1985). Stated simply, Zipf's law argues that the frequency of utterances should be reciprocally related to their frequency rank—the tenth most common utterance (word in a human language, call syntax in the chick-a-dee call) should occur with 1/10th the frequency of the most common utterance. Human language meets this criterion, but black-capped chick-a-dee calls do not.

A notes D notes

A notes D notes

Figure 13.3 Survivorship plots of A notes and D notes of the Carolina chickadees. These are based on a sample of 2153 calls. The triangles represent calls in the sample, the line is the predicted survivorship from a semi-Markovian model, based on the transitions between same-type notes (e.g. in the left panel, the transition probability from A to another A note).

Figure 13.3 Survivorship plots of A notes and D notes of the Carolina chickadees. These are based on a sample of 2153 calls. The triangles represent calls in the sample, the line is the predicted survivorship from a semi-Markovian model, based on the transitions between same-type notes (e.g. in the left panel, the transition probability from A to another A note).

However, black-capped calls do fit a broader form suggested by Mandelbrot (Hailman et al. 1985). Interestingly, our Carolina chickadees do not fit Mandelbrot's function (Fig. 13.4), at least based on a least-squares best fit of the data. Nonetheless, the general increase in cumulative call types with an increase in the number of calls sampled indicates that the call system is generative, or open-ended.

As Hailman et al. (1987) showed in black-capped chickadees, the probability that a call ends (instead of continuing with a new note) increases with an increase in the number of A and C notes (Table 13.1). In Carolina chickadees, longer strings of A notes (eight-ten) are more likely to transition to C notes whereas shorter strings (one-five) are as likely to transition to B or D notes. B notes almost always transition into D notes or end the call.

We need to add a caveat that the differences between black-capped and Carolina chickadees could result from differences in the field recording contexts. Assuming that this caveat does not generate a bias in our data sets, the results suggest that Carolina chickadees use long strings of introductory notes in different ways (with different meanings?) than black-capped chickadees. It would be instructive to perform this analysis on mountain and Mexican chickadees, given the species differences reported earlier. It would also be o : e '

Figure 13.4 Percent use of chick-a-dee variants as a function the frequency rank (1 = most common). Triangles are data from our set of 2153 calls of Carolina chickadees. Line represents the best fit line: p = i(r+k)~s, where p = percent use, r = rank, and i, k, and s are fit constants (see Hailman etal. 1985).

instructive to see this analysis done on European tits, particularly the marsh and willow tits, which are closely related to the North American chickadees (Gill et al. 2005).

13.2.5.2 Context

Studies of correlations between chick-a-dee variants and behavior are critical because they give us some insight into the potential for syntactically-mediated information transfer. Results to date clearly o : e '

1.0 10.0 100.0 Frequency rank
Table 13.1 Transition probabilities for strings of A, B and C notes in chick-a-dee calls of Carolina chickadees; in each case, the number of elements in a string of same-type notes is given, followed by the probability that the string ends in another note type or silence (A0, B0, or C0)

AA

AB

AC

AD

A0

BB

BC

BD

B0

CC

CD

C0

1

0.11

0.32

0.36

0.21

1

0.06

0.82

0.12

1

0.81

0.19

2

0.20

0.20

0.29

0.31

2

0.5

0

0.5

2

0.77

0.23

3

0.10

0.22

0.29

0.39

3

0

0.5

0.5

3

0.48

0.52

4

0.08

0.26

0.08

0.58

4

0.33

0.67

5

0.06

0.14

0.28

0.52

5

0.63

0.37

6

0

0.11

0.05

0.84

6

0.36

0.64

7

0

0.11

0.35

0.54

7

0.43

0.57

8

0

0.14

0

0.86

8

0.50

0.50

9

0

0.17

0

0.83

10

0

0.20

0

0.80

11

0

0

1.0

0

13-17

0

0

0

1.0

demonstrate that different chick-a-dee variants are given under different conditions.

The first extensive, though qualitative, study of this type was published by Smith (1972) on Carolina chickadees. Smith suggested that different note types had different meanings. For example, chip (a C-variant) notes are aggressive calls whereas chick notes (another C variant) are non-aggressive, often heard around feeding stations. Haftorn's (1993) study of willow tits demonstrated similar meaning in the C note, with one variant characteristic of alarm calls and another typically embedded in less aggressive calls.

Ficken et al. (1994) suggested that A notes are given by Mexican chickadees moving in space, whereas C notes are given in response to a disturbing stimulus (e.g. when mobbing a screech owl tape) or when birds change directions. D notes tend to be given by perched birds. These trends for A and D notes are similar to those reported by Smith (1972) for Carolina chickadees. Gaddis (1985) found that chick-a-dee variants were context specific in mountain chickadees. For example B notes are given when birds leave food; A notes are given when birds fly up, and [A][D] strings are given in flocks.

There are at least three studies of the syntactical cues given in black-capped chickadee mobbing calls. Hurd (1996) suggested that mobbing calls had more introductory notes than non-mobbing calls but the same number of C and D notes. Baker and Becker (2002) showed a similar pattern, with more B notes and fewer A notes given under more immediate risk (1 m vs. 6 m from a stuffed prairie falcon), but no difference in the number of C or D notes. In contrast, Templeton et al. (2005) showed that the number of D notes correlated strongly with the intensity of risk represented by different species of predators. It is not clear why these results are so different, although there were considerable methodological differences across the three studies.

In addition to syntactical cues, some cues may be given by call rate itself, with higher rates indicating more intense conditions. Black-capped chickadee mobbing calls, for example, tend to be given at higher rates under more immediate risk (Baker and Becker 2002). Carolina chickadee chick-a-dee call rates are higher when the birds are light-weight and hungry than when they are heavy or sated (Lucas et al. 1999)—chick-a-dee calls may encode information about signaler physiological condition. We (K. Bledsoe and J. Lucas, unpublished data) have limited data on two Carolina chickadees that indicate just this: D note fundamental frequency and duration correlated strongly with changes in corticosterone levels. These results are consistent with Owings and Morton's (1998) assessment/ management model of communication, although clearly we need more extensive studies of this aspect of the chick-a-dee complex.

13.2.5.3 Playback studies

We can use playback experiments to test implications about syntactical information derived from field (or laboratory) observational studies. Here we focus on recent field-based playback studies (Sturdy et al. review laboratory-based studies of perception in Chapter 10). Freeberg and Lucas (2002) proposed that the C-note (chick variant) was food related, based on preliminary field observations. They tested this by broadcasting either C-rich or D-rich chick-a-dee calls at a temporary seed stand. Consistent with the hypothesis, birds tended to come to the stand and take seeds in response to C-rich calls but never took a seed in response to D-rich calls. Moreover, the rate of chick-a-dee calling was significantly higher in playbacks with C-rich calls than those with D-rich calls. There are at least two alternative explanations for this result. One is that the C note is indeed a food-associated note. The second is that D-rich calls are aggressive calls that elicit an aggressive reaction by receivers (and, by comparison, C-rich calls are non-aggressive). While we cannot distinguish these alternatives with this experimental design, the results indicate that chick-a-dee variants vary in their meaning to receivers.

Templeton et al. (2005) tested the relative function of black-capped chickadee mobbing calls. They found that the number of D notes was negatively correlated with predator wingspan and body length (smaller, more dangerous, predators elicit more D notes). Smaller predators elicited D notes with a narrower band width and more narrowly spaced overtones. Templeton et al. (2005) used playbacks of the mobbing calls in the absence of predators as an important test of the proposed information embodied in the calls. The birds gave responses appropriate to the predator that was being mobbed when the calls were first recorded, suggesting the calls conveyed some quantitative index of predation risk.

Finally, Clucas et al. (2004) monitored Carolina chickadees' responses to artificially constructed calls that varied in both note composition (AAAACCCC, AAAADDDD, and CCCCDDDD) and note ordering (AAAACCCC vs. CACACACA, CCCCDDDD vs. DCDCDCDC). The playback was repeated in two seasons, spring and fall/winter, to test for the potential role of seasonal context on receiver's responses to the calls. The experiment tested whether chickadees respond to manipulation of the two components of syntactical organization, note composition and note order, but did not test any specific hypotheses about the meaning of chick-a-dee variants. The birds showed no differential response to calls with atypical note ordering that varied in note composition (CACACACA vs. DCDCDCDC). Birds did, however, distinguish between calls that varied in note composition if the note order followed the species-typical A-B-C-D ordering. Season and social context also affected the birds' response to the playbacks. The number of introductory notes in the responding birds' chick-a-dee calls was affected by note composition in the fall but not in the spring: AAAACCCC playback calls elicited A/B-rich responses, and D-rich playback calls elicited C-rich responses. But this latter response was evident only when white-breasted nuthatches and tufted titmice (two dominant het-erospecifics) were present. In the fall, the number of D notes in the call was also positively correlated with the number of birds responding, but this relationship was positive only when heterospecifics were not present. Only the number of birds responding affected the number of D notes in the spring [note: Table 2 in Clucas et al. (2004) is correct, but the discussion of this pattern in the text of that paper and Figure 8 are not correct].

Our results suggest that both note ordering and note composition affect the information content of the chick-a-dee complex. In addition, context (season and presence of dominant heterospecifics) matters, as predicted by Hailman et al. 1985 (also see Leger 1993; Marler and Evans 1996).

Was this article helpful?

0 0

Post a comment