volume 7
april 2004

Semantic shifts in Beatles' chord progressions

On the perception of shifts in song contexts induced by chords

by Juul Mulder

Many music text books stubbornly hold to the thesis, that Minor scales sound sad whereas Major scales, in contrast, have a more cheerful ring. This observation, though, does not to apply to popular music at large. At least, that is the conclusion derived by Ger Tillekens (1998) from his analysis of the chord transitions in the Beatles' songs. These transitions, he argues, rather reflect their lyrics by referring to changing contexts of conversation. However, can and do listeners actually perceive chord transitions that way? Using an experimental design, Juul Mulder tried to answer this question. Here we reprint her results.

The semantic meaning of the Beatles' chords. Sounding loud and clear, a new form of popular music, beat music, made itself heard in the 1960s. At the same time youth propagated more egalitarian and informal ways of communication, modelled upon the peer group, as a new standard for social interaction. According to Tillekens (1998; 2001) both these changes are closely tied together. His study The Sound of the Beatles builds up quite an argument for a direct relationship between the new musical forms of the decade — exemplified by the music of the Beatles — and the communication codes of the peer group, characterized by an open and almost permanent negotiation of feelings and opinions. The harmonic progressions typically used by beat musicians, Tillekens argues, represent semantic shifts within the egalitarian context of peer group conversation. Moreover, because these chord sequences were very flexible, the songs acquainted their listeners with a new and also more flexible way of communication.

Combining the tools of sociology and musicology, Tillekens offers convincing theoretical support for a direct relationship between these two types of increased flexibility (Breeuwsma, 1999). He, however, does not provide any direct empirical evidence that the young people of those days actually were capable of perceiving harmonies in a way as to have an influence on their way of thinking. The missing link in his study, thus, is a psychological one. Lacking is an indication that listeners could and can actually perceive these chord transitions in the proposed way. Can harmonies in popular music really be perceived as representations of specific conversational contexts? The present experimental study is meant as a first step in answering this question.

Chords and conversational contexts. In the compositions of the Beatles, Tillekens finds a drastic extension of the conventional chord material of popular music. Moreover, all these chords are combined in surprising and innovative progressions. This increased harmonic flexibility parallels, he argues, the flexible contextual shifts typical for peer group communication. Chords and words are semantically connected, because both refer to the same conversational contexts. In this way the songs of the Beatles could and did offer a new extended model for a more egalitarian and open conversation among peers.

Based on his analysis of the first 46 Beatles' songs, Tillekens proposes that in the songs of the Beatles the conventional three basic chords known as the tonic (I), the subdominant (IV) and the dominant (V), can be replaced almost at will by approximately five other types of chords. These categories comprise the Relative Minors (vi, ii, iii), the Relative Majors (VI, II, III), the Parallel Minors (i, iv, v), Parallel Majors (flat-III, flat-VI, flat-VII) and, last but not least, the Seventh Chords (I⁷, IV⁷, V⁷). The substitution of these types of chords follows a diagonal structure as can be seen in Figure 1. In the key of C Major, for instance, the C chord (I) can be replaced up along the diagonal lines by A minor (vi), by A Major (VI), and by F# minor (#iv). Downwards the C chord can be replaced diagonally by C minor (i) and, one step further down, by E-flat Major (flat-III). For the subdominant and dominant — in this case the F (IV) and G Major (V) chords — a similar logic applies.

Figure 1: The extended chord material of the early Beatles' songs
(cfr. Tillekens, 1998)

Tillekens' main thesis is that the diagonal substitution of chords is steered by shifts in the semantic contexts of the song's lyrics. Following Harré (1983), these contexts can be delineated along two orthogonal dimensions — see Figure 2. The first dimension is called "display" and has as its extremes private and public space as the locations for conversation. The second dimension relates to the speaker's willingness and ability to act upon his words and therefore is called "realization." This dimension starts at one extreme with inner monologue, and transforms the character of an utterance through personal dialogue to the other extreme of making a public statement. The crossing of the two dimensions represents the domain of the peer group, implying the two dimensions should be regarded as continua.

Figure 2: A two-dimensional representation of conversational contexts
(cfr. Harré, 1983)

According to Harré (1983), in conversations each speaker takes up an argumentation position in this two-dimensional space as a speaking actor, changing this position as the dialogue develops. The first, vertical dimension "display," indicates the particular space — private or public — and with that for whom the song is intended. The language used and the tone in which it is voiced, for instance, may be informal or formal, depending on the fact whether the message is uttered in the private confines of home or in a public place. The second, horizontal dimension "realization," concerns the actor's warrantability. Overt statements can be checked by bystanders and require certainty, while inner monologues lend themselves more easily to the expression of doubts and wishes. Realization, thus, can be described in terms of opposite adjectives like as "inner" versus "overt, "uncontrolled" versus "uncontrolled," "uncertain" versus "certain," "primary" versus "secondary," or "uncensored" versus "censored." Taken together both dimensions constitute an opposition of "impolite" versus "polite" utterances.

Situating chords in their contexts. Through their semantic potential, chords acquire their position in this two-dimensional matrix of conversation depending on their accompanying function — see Figure 3. That function can be the accompanying of the more formal language of the public domain; of the personal utterances within the private domain; or, in between, of the egalitarian domain of the peer group. Seventh chords accompany statements for the entire world to hear. So do basic chords, especially the dominant, but here the utterances have a less formal character. Transitions to Parallel Minors induce a more confidential tone, like when sharing a secret. The transition to Parallel Majors, also known as Neapolitan chords, mainly accompanies a shift towards the honest expression of deeply felt emotions. Relative Minors are used to accompany situations with people you know, but with the private character of an inner monologue. Relative Majors come with uncensored, public expressions of feelings.

Figure 3: Types of chords representing conversational contexts
(cfr. Tillekens, 1998)

Music perception is a difficult research area, especially when it comes to chord perception. The subject has been investigated in several ways, but mostly in relation to perceived emotional meaning (e.g. Sloboda, 1991; Nielzén et al., 1981) or in relation to music-theoretical notions, like the hierarchical organization of the Western tonal music system in which chords are viewed as the intermediate level in the hierarchy from bottom-up input, i.e. notes, to the complex phenomenon of key perception — strongly top-down (Bahrucha, 1987). This traditional view is challenged by Povel et al. (1993) who found the relationship between the melody of a piece of music and the harmonic accompaniment to be one of relative independency, with changing harmonic accompaniment — without a changing key — not interfering with melody recognition. The authors conclude that melodic and harmonic aspects of music give rise to distinct perceptual effects.

It is important to notice that for a listener to be able to judge the significance of chords, a tonal context needs to be established first. Chords presented in isolation would not mean a thing, neither in a musical nor in a semantic sense. A key needs to be established for a chord to have any meaning,

"because a key establishes a kind of hierarchy on the sets of tones and chords in terms of more or less related to the prevailing tonality, with less related chords being perceived as less stable and closely related chords as stable (Krumhansl, 1990).

This knowledge does not have to be explicitly available as it is to musical experts. Non-musicians also show evidence of having internalized key recognition through mere exposure to Western tonal music as was indicated in their brain electrical responses — known in the psycho-physiological literature as event-related potentials; ERPS — to stimuli in keeping with the music-theoretical and phenomenal stability of chords (Koelsch et al., 2000). Even starlings (Sturnus Vulgaris) show perceptual sensitivity to chord-based spectral structures constructed on the basis of principles of musical intervals (Hulse et al., 1995), though the data could also be explained in terms of the distinction "consonance versus dissonance."

Method of the study. The intention of our experiment was to investigate whether harmonic shifts can be perceived as shifts in the semantic contexts of a song excerpt. To this end we presented progressions of two and three chords to a sample of 40 subjects. The first chord was meant to induce a tonal context for the next ones, which we asked the subjects to rate on both semantic dimensions. The stimuli contained a minimal amount of musical information. The song fragments with the relevant chord transitions were reduced to chords played on a piano, each lasting only a couple of seconds. The approach taken thus can be called highly reductionistic. The research therefore not only aimed at a validation of the theory, but also tried to explore whether this approach was viable.

Stimuli. For the experiment 54 chord progressions were chosen from the early Beatles' songs and transposed into one of three tonal keys. The selected tonal keys used were C, E, and G. Of each of these three musical scales three different chord types were recorded, namely the minor, Major and Major Seventh chord types. Each chord was prepared in two different octaves. A Yamaha SWXG60 synthesizer was used for the chord sampling. The recordings were saved in midi-files, each containing the notes, volume and lengths of notes. A grand piano sound was chosen to present the chords. The chords sounded for only two seconds. Next, these files were transformed into wav-format and subsequently combined to form matching two-chord and three-chord sequences. These files, in turn, were converted to RealAudio as RealAudio Player was used as an utility for playing the files. In this way, all in all, 108 files were constructed: 54 two-chord sequences and the same amount of matching three-chord sequences.

Sample of a two-chord sequence

Sample of a three-chord sequence

The number of 54 stimuli, by the way, was derived from the number of possible combinations of shifts over the two dimensions of "display" and "realization." These combinations add up to nine different types of harmonic shifts — see Table 1. As we chose to let each subject rate six matching pairs of two-chord and three-chord progressions for each of these nine categories, we ended up with 54 blocks of stimuli.

		Display
		+	0	-

	+	A	B	C
Realization	0	D	E	F
	-	G	H	I

Table 1: Nine categories (A to I) of semantic shifts

Subjects. Of the 40 participants half were female and half were male. The average age was 21.5 years, with a range of 18 to 32 years (SD = 2.9). Four persons were left-handed. In the call for participation for this experiment, people were requested with a good feel, a musical intuition, and a general liking for pop music. Active musical abilities, however, were no precondition. The selection criterion for participation, therefore, was not an objective one. The call attracted people who were appealed by the topic under investigation. Before the actual experiment the participants were asked a couple of questions concerning their musical abilities as we wanted to know if this characteristic would influence the outcomes. No other background variables were taken into account, because, at least concerning the perception of emotional meaning in music, there is hardly any influence found of age (Terwogt & Van Grinsven, 1988), gender or even expertise (Robazza et al., 1994), with musical experts and non-experts ascribing similar emotional reactions to pieces of music. All subjects were asked to rate their musical sensitivity on a scale from 1 to 5, ranging from "lousy" [sic!] to "excellent." The distribution of these characteristics is shown in Table 2 below.

Musical Sensitivity		N of Subjects

1.	Lousy	-
2.	Bad	1
3.	In between	10
4.	Good	26
5.	Excellent	3

		Total: 40

Table 2: Self-ratings of musical sensitivity

We also asked the subjects if they were actively involved in singing or playing an instrument and if they could read score. Of the 40 respondents, 25 were actively involved in playing music and 15 were not. Not surprisingly, these characteristics proved to be related. The correlation between musical sensitivity and musical involvement came out on .33 (p < .05), the correlation between musical sensitivity and reading score was .38 (p < .05), and the correlation between musical involvement and the capacity of reading score was .41 (p < .01).

Procedure. The experiment was run on a conventional multimedia PC by way of a series of interactive Internet pages, each presenting six blocks of chord sequences to the subjects. Participants could work at their own pace because they could hear every next chord progression by clicking on a "play" icon. Chord progressions were offered in blocks of two matching sequences. The first sequence comprised the first two chords, with the first chord representing the tonic. The second chord was the first item to be rated. The second sequence comprised the same two chords plus an added third one. Again the final chord had to be rated by the subjects. All 54 pairs were run twice, first for the "display" dimension and then again, to avoid semantic confusion, separately for the "realization" dimension. On the Internet pages the dimensions were represented by rating scales with the points -1; -0.5; 0; 0.5; 1. The extremes "private" (display) and "inner monologue" (realization) were indexed by -1, and the other extremes, "public" (display) and "overt conversation" (realization), were indexed by +1. The three points in-between were employed to convey a sense of continuity of the dimensions and an idea of a "neutral" centre — look here for a sample page.

Instructions. Because of the artificiality of the task an example was given at the start to convey the meaning of the dimensions to our subjects. When they said they had gotten the idea, the experiment was started. Subjects were asked to simply listen well to the first chord as it would provide the tonal context, and they were told to rate the second and third chord on both dimensions. They were instructed that a harmonic shift did not necessarily always imply a semantic shift — the new chord could still belong to the same chord type, e.g. the dominant or subdominant. The instructions were given verbally so the respondents could ask questions to the test leader if and when needed. To explain the dimension "display" the subjects were told to concentrate on the sound of the presented chord and to try to think of what type of space comes to their mind: a private one or a public one — in a relative way; it is not a dichotomy. In regard to the dimension "realization" the subjects were asked to assess the degree of certainty with which someone professes his/her feelings and opinions. In an inner monologue all doubts can come forward, because no one can judge you by your words or rather thoughts. In the anonymous company of strangers, on the other hand, you have to be certain of what you have to say. And, in talking with your peers you have to mix openness with social appropriateness. To pinpoint this difference the test leader kept to the adjectives "inner" versus "open" or "overt" to denote the character of the conversation.

Data analysis. The data analysis was done from two different points of view. The first approach follows directly from the original research setup. Here we analyzed the semantic shifts by checking if the subjects rated the chord sequences in the same direction as predicted by the theory. To this end, difference scores were calculated for each presented chord sequence by subtracting the rating of the second chord from the rating of the third one. Next, we looked at the number of correct scores for each dimension. As the results of this first approach proved difficult to interpret, a second approach was adopted. Again we computed difference scores. Combining the results of both dimensions, the scores now were ordered according to the chord type of the final chord. For each chord type the internal consistency of the scores was tested by means of a reliability analysis to check their scalability. Items that did not fit in were removed. The remaining ones were scaled and, next, the resulting scales were put to the test by a correlation analysis and an exploratory factor analysis, meant to visualize their mutual relations. In the first as well as the second approach the analysis was performed separately for the rating of the second chord and the rating of the third chord.

Results of the first analysis. Our first analysis was meant to investigate the theory in terms of shifts on the dimensions of display and realization. To see if people were able to perceive the correct shifts in semantic contexts, difference scores were calculated by subtracting the rating of the second chord from the rating of the third chord. Next, all scores were rated according to their correspondence to the theoretical positions of the sequences. To be rated as correct, a shift only had to be in the right direction, absolute positioning was not looked at. Table 3 offers an example for the theoretical values of a chord sequence taken from the first measures of the verse of "I'm A Loser," i.e. G -» Dm -» F -» G; or I -» iv -» flat-VII -» I. For the dimension of display the difference between the third en second chord here is zero; for the dimension of realization the difference between the same chords is minus two or negative.

Semantic shift

Chord Progression

Display

Realization

-»

-1

+0.5

-»

-1

-»

-1

-»

-1

Difference:

-2

Table 3: A three-chord sequence and its semantic shifts

Of the dimension "display" on average 17.2 (out of 54, i.e. 32%) pairs of chord sequences were judged correctly. During the test one particular webpage with stimuli of the dimension "realization" got lost on the net. The responses of that page somehow did not get stored. Therefore, for this dimension six pairs of chord sequences were dropped, making the total amount of items for this dimension 48. For the dimension "realization" the average of correct pairs came out on 14.2 (out of 48, i.e. 30%). Overall, one third of the items was judged in accordance with the theoretical expectations. That is about the amount of correct ratings that could have been expected if the subjects had rated the shifts at random. Some categories of semantic shifts, though, did better and others did worse: especially the no-shift categories proved to be problematic — see Table 4. The failure of this category may be due to a response bias, as there were more shifts than no-shifts.

	Combined		Mean number of
Type of	movement on:		correct ratings on:
shift	Display	Realization	Display	Realization

A	+	+	2.9	2.2
B	0	+	1.0	1.9
C	-	+	1.6	1.8
D	+	0	2.3	0.9
E	0	0	1.1	1.0
F	-	0	2.6	0.8
G	+	-	2.0	1.6
H	0	-	0.9	1.8
I	-	-	2.8	2.3

		Total:	17.2	14.2

Table 4: Mean number of correct ratings for nine categories of semantic shifts on the dimensions of display (N = 54) and realization (N = 48)

The results were rather disappointing. Some subjects, though, performed slightly better than others. The independent variables, measured at the start of the experiment, proved to be discriminative on the outcomes of the shift ratings. Added up for both dimensions, our subjects rated 102 blocks. Subjects indicating their musical sensitivity to be 5 on an ascending scale from 1 to 5, on average judged 42 items in line with the theory. Taken together subjects who gave themselves a 4 or a 5, scored on average 40 and people judging themselves to be less than 4 on the scale scored 38.5 on average. People actively or not-actively involved with music, i.e. performers or non-performers, did equally well. This was not discriminative. Nor was the ability to read musical notation. However, the combination of these three characteristics proved to be discriminative: people able to read music scores and actively involved in making music and having a musical sensitivity of 4 or 5 on the scale, on average scored 41 versus 37 correct answers out of 102.

Results of the second analysis. Having come this far, we decided on a second analysis to look into the consistency of the ratings of the chord sequences for each chord type. To this end the subject scores on both dimensions now were taken together in one and the same item pool. After having converted the initial scores to values ranging from 1 to 5, we again computed difference scores for the rating of the third chord. Next, the items were categorized according to the type of their final chord in each of the six chord types as shown in Figure 3. Table 5 shows some examples of chord sequences and the types of their final chords.

Scales

Two-chord Sequence

Three-chord Sequence

Seventh Chords

-»

Relative Majors

-»

Basic Chords

-»

Relative Minors

-»

Parallel Minors

-»

Parallel Majors

-»

Table 5: Some examples of chord sequences with different types of final chords

For each type of chords, the items were examined on their scalability by computing Cronbach's Alpha. For some types of chords many items had to be removed, but others did rather well. For the two-chord sequences the items of the dimension of realization performed best on the reliability scaling. Of the six categories of chord progressions the Parallel Majors here were perceived least consistent: 10 out of 17 items had to be removed and the resulting Alpha still was rather low. The Basic Chords and the Seventh Chords, however, did very well — see Table 6.

Two-chord Scales			Items		N	Minimum	Maximum	Mean	Std. Dev.	Alpha

1.	Seventh Chords	16	out of	16	36	1.4	4.3	3.2	0.7	.90
2.	Relative Majors	6	out of	11	35	1.3	4.7	3.0	0.9	.84
3.	Basic Chords	20	out of	20	38	2.7	4.5	3.4	0.5	.80
4.	Relative Minors	14	out of	21	36	1.9	3.9	2.8	0.5	.69
5.	Parallel Minors	8	out of	17	39	1.6	4.3	2.7	0.7	.76
6.	Parallel Majors	7	out of	17	40	2.1	5.0	3.0	0.5	.56

Table 6: Some statistics for the two-chord rating scales

For the three-chord sequences the items of both dimensions performed equally well in contributing to the scales' reliability. Also, overall more items were left: 75 instead of 71 out of 102. Here the basic chords were perceived least consistent. With the Seventh Chords, the Parallel Minors now proved to provide the best scalable items — see Table 7.

Three-chord Scales			Items		N	Minimum	Maximum	Mean	Std. Dev.	Alpha

1.	Seventh Chords	18	out of	20	37	-1.7	1.2	0.0	0.7	.84
2.	Relative Majors	13	out of	20	39	-1.1	1.8	0.4	0.8	.78
3.	Basic Chords	4	out of	11	38	-1.5	3.0	0.4	0.9	.70
4.	Relative Minors	8	out of	12	39	-2.3	1.3	0.0	0.7	.63
5.	Parallel Minors	20	out of	20	33	-1.4	0.6	-0.7	0.7	.81
6.	Parallel Majors	12	out of	19	37	-1.4	2.0	0.5	0.6	.69

Table 7: Some statistics for the three-chord rating scales

Next, we computed the values for the two-chord and three-chord rating scales by adding up the relevant scores of the remaining items for each subject. By combining the scores for both dimensions — realization as well as display — it may seem that we have lost information about the dimensions themselves. The necessary information, however, survives in the variation and still can be inferred from the correlations between the scales. We will visualize these relations in the next paragraph, but first we will present the correlation matrixes for the two-chord — see Table 8 — and three-chord rating scales — see Table 9.

	1. Seventh		2. Relative		3. Basic		4. Relative		5. Parallel		6. Parallel
Two-chord Scales	Chords		Majors		Chords		Minors		Minors		Majors

1. Seventh Chords	(1.00	)
2. Relative Majors	.54	**	(1.00	)
3. Basic Chords	—		—		(1.00	)
4. Relative Minors	-.46	*	-.59	**	—		(1.00	)
5. Parallel Minors	—		—		—		.47	**	(1.00	)
6. Parallel Majors	—		.35	*	—		—		—		(1.00	)

Musical Sensitivity	—		—		.39	*	—		—		.37	*
Musical Involvement	—		—		.43	**	—		—		—

Table 8: Statistically significant correlations for the two-chord rating scales
(*= p < .05, **= p < .01)

For the two-chord rating scales, the Basic Chords (Scale 3) proved to correlate positively with musical sensitivity and musical involvement. This finding implies that this type of chord is perceived as more "open" and referring to the "public" sphere by subjects having a feeling for music and/or being actively involved with music. For musically sensitive subjects the same goes in respect to the Parallel Majors.

	1. Seventh		2. Relative		3. Basic		4. Relative		5. Parallel		6. Parallel
Three-chord Scales	Chords		Majors		Chords		Minors		Minors		Majors

1. Seventh Chords	(1.00	)
2. Relative Majors	.51	**	(1.00	)
3. Basic Chords	-.67	**	-.38	**	(1.00	)
4. Relative Minors	—		—		—		(1.00	)
5. Parallel Minors	-.44	**	-.48	**	.47	**	—		(1.00	)
6. Parallel Majors	—		.41	*	—		—		—		(1.00	)

Left-handedness	—		—		—		-.45	**	—		—

Table 9: Statistically significant correlations for the three-chord rating scales
(* = p < .05; ** = p < .01)

For three-chord rating scales no significant correlations with the background variables were found, except for an unexpected finding. The Relative Minors (Scale 4) proved to be correlated negatively with left-handedness. Compared to their right-handed peers, lefties seem to perceive Relative Minors more strongly as referring to inner monologue and private space.

Plotting chord types. To visualize the relations between the rating scales, an exploratory component analysis was performed, both on an ordinal (Princals) and an interval (Factor) level. Both analyses produced similar plots. We here present the results of the factor analysis, thereby assuming that the distances between the chord rating scales can be measured on the interval level. First, replicating Figure 3, we once again, show the relations we would have expected theoretically, now rotated to conform to the outcomes of the factor plots — see Figure 4.

Figure 4: Theoretical positions of the six types of chords in two-dimensional space

A two-factor solution explained 61% of the variance. The diagram of Figure 5 shows some resemblance with the theoretical model (Figure 4), though there are some deviations or rather reversals. Most conspicuous is the reversal of the Basic Chords (scale 3) and the Relative Minors (scale 4) versus the Parallel Majors (scale 6) and the Parallel Minors (scale 5) on the display dimension. Moreover, in respect to both last chords types, there is also a reversal Parallel Majors (scale 6) and the Parallel Minors (scale 5) on the realization dimension.

Figure 5: Two-dimensional factor plot for the two-chord rating scales

For the three-chord sequences, the explained variance of the factor analysis added up to 66% for the two dimensions. The better performance of this analysis possibly can be explained by the fact, that the distances between the second and the third chords overall were greater than those between the second and first chords.

Figure 6: Two-dimensional factor plot for the three-chord rating scales

As can be seen in Figure 6, the reversal between the Parallel Minors (scale 5) and the Parallel Majors (scale 6) now has been undone and the diagram does resemble the theoretical model more clearly. There are, however, still some deviations. The position of the Basic Chords (scale 3) still is too much inclined towards the "public" pole of the display dimension, as are the Relative Minors (scale 4) towards the "open" pole of the realization dimension. The Parallel Majors (scale 56) now are positioned at the wrong end of the display axis.

Discussion. Our first analysis, looking at the number of correctly perceived shifts on the two dimensions, yielded few results. The overall performance of the subjects was rather faulty. Looking at the perception of semantic shifts, we found that approximately 30% of the responses was in accord with our theoretical expectations. No interpretable patterns were found. The second analysis did better, a least indicating that people can perceive and interpret harmonic shifts as shifts in semantic meaning. The category of Seventh Chords was judged almost perfectly according to theory, i.e. as accompanying public statements in a public space. However, there were some puzzling reversals. Some of those may be due to the way in which the subjects were instructed. To denote the character of the conversation the test leader used adjectives like "inner" versus "open" or "overt." Possibly an alternative meaning of "open" was implicitly used by subjects. In the experiment "open" was meant to denote a more censored utterance, made in public so bystanders will judge the words more severely on their consequences. Our subjects, in reverse, may have interpreted "open" in the sense of more honest, i.e. less censored.

Yet another dimension may have been interfering with the perception and interpretation of the chord sequences by the subjects. Apart from "display" and "realization" Tillekens (1998) discerns yet another, more primary dimension, indicating the relations between the subdominant, the tonic and the dominant. This dimension, called by the name of "agency," refers to the steps taken by a song's proponent in taking actions, respectively thinking, deciding and acting. Here we also find associations with the oppositions between "inner" and "overt" and between "private" and "public." In a follow-up study, the explanation of the dimensions clearly will have to be more elaborate. To detect which terms and concepts are most fit to denote the perception of shifts in contexts, the next logical step would be to apply a semantic differential method to the perception of specific types of chords. This has been done before by Nielzén and Cesarec (1981), though these authors used adjectives denoting emotions, and did not focus explicitly on the conversational contexts of music. Their analysis came up with three dimensions: tension, gaiety and attraction. These are not the type of dimensions relevant to this study. The technique itself, however, has proven useful in trying to discern semantics in music.

Despite the huge reduction in song information, participants performed fairly well in perceiving the semantic potential of harmonies. The finding offers some tentative empirical support for the claim made by Tillekens that the Beatles' songs — with their distinctive combinations of chord progressions and lyrics — helped to bring about communicative changes in their listeners. This does not fit in with common sense thinking about harmonies, even by our subjects. During the experiment they were asked what they thought of the theory itself. Usually they said that it was not inconceivable. At the same time they professed of conceiving of harmonies as being related to emotional states, with Major chords fitting well with happy passages and minor chords with sad ones. This is also what researchers in music psychology in general are focusing on: music and perceived emotion. Sloboda (1991), for example, asked people to select musical pieces meaningful to themselves, and to indicate where in the piece which emotional reaction was evoked — in physical terms such as shivers down the spine, laughter, etcetera. North and Hargreaves (1997) used non-vocal excerpts from pop songs to investigate whether liking and arousal potential of the excerpts could predict a variety of perceived emotions in these excerpts. However, as this experiment has shown, there is more to find in harmonies than just emotions.

References

Bharucha, Jamshed J. (1987), "Music cognition and perceptual facilitation. A connectionist framework." In: Music Perception, 5, 1-30.
Breeuwsma, Gerrit (1999), "The greatest of them all. Over jeugdcultuur, Beatles en de uitvinding van plezier." In: Psychologie en Maatschappij, 23, 2, 166-178.
Butler, David, and Helen Brown, (1984), "Tonal structure versus function. Studies of the recognition of harmonic motion." In: Music Perception, 2, 6-24.
Harré, Rom (1983), Personal being. A theory for individual psychology. Oxford: Blackwell.
Hulse, Steward H., Daniel J. Bernard and Richard F. Braaten (1995), "Auditory discrimination of chord-based spectral structures by European starlings (Sturnus Vulgaris)." In: Journal of Experimental Psychology, 124, 4, 409-423.
Krumhansl, Carol L. (1990), Cognitive foundations of musical pitch. Oxford: Oxford University Press.
Kölsch, Stefan, Thom C. Gunter, Angela D. Friederici and Erich Schröger (2000), "Brain indices of music processing. "Non-musicians" are musical." In: Journal of Cognitive Neuroscience, 12, 520-541.
Nielzén, Sören, and Zvonimir Cesarec (1981), "On the perception of emotional meaning in music." In: Psychology of Music, 9, 17-31.
North, Adrian C., and David J. Hargreaves (1997), "Liking, arousal potential, and the emotions expressed by music." In: Scandinavian Journal of Psychology, 38, 45-53.
Povel, Dirk-Jan, and René van Egmond (1993), "The function of accompanying chords in the recognition of melodic fragments." In: Music Perception, 11, 2, 101-115.
Robazza, Claudio, Cristina Macaluso and Valentina D'Urso (1994), "Emotional reactions to music by gender, age, and expertise." In: Perceptual and Motor Skills, 79, 2, 939-944.
Sloboda, John A. (1991), "Music structure and emotional response. Some empirical findings." In: Psychology of Music, 19, 110-120.
Temperley, David (1997), "An algorithm for harmonic analysis." In: Music Perception, 15, 31-68.
Terwogt, Mark Meerum, and Flora van Grinsven (1988), "Recognition of emotions in music by children and adults." In: Perceptual and Motor Skills, 67, 697-698.
Tillekens, Ger (1998), Het geluid van de Beatles. Een muzieksociologische studie. Amsterdam: Het Spinhuis.
Tillekens, Ger (2001), "Words and chords. The semantic shifts of the Beatles' chords." In: Soundscapes, 3, Summer.

This article is a slightly adapted version of: Mulder, Juul (2001), "Semantic shifts in Beatles' chord progressions. On the perception of shifts in song contexts induced by chords." In: Yrjö Heinonen, Marcus Heuger, Sheila Whiteley, Terhi Nurmesjärvi and Jouni Koskimäki (eds.), Beatlestudies 3. Proceedings of the Beatles 2000 conference. Jyväskylä: University of Jyväskylä (Department of Music, Research Reports 23), 2001, 113-128.