JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

DELAYED MATCHING-TO-SAMPLE PERFORMANCE OF HENS:EFFECTS OF SAMPLE DURATION AND RESPONSE REQUIREMENTS

DURING THE SAMPLE

T. M. FOSTER, W. TEMPLE, C. MACKENZIE,L. R. DEMELLO, AND A. POLING

THE UNIVERSITY OF WAIKATO, THE UNIVERSITY COLLEGE OF BALLARAT,AND WESTERN MICHIGAN UNIVERSITY

Six domestic hens were trained under a delayed matching-to-sample procedure with red and greenkeylights as sample and comparison stimuli and a 1.5-s delay interval. The hens were trained to stoppecking the sample stimuli when a tone sounded. Duration of the sample stimuli (2 to 10 s) andthe number of pecks required on the key on which these stimuli were presented (O to 10) werealtered across conditions. Both the response requirement on the sample key and the duration ofsample presentations affected accuracy. These findings are in agreement with those of earlier studiesusing other species and somewhat different procedures.

Key words: delayed matching to sample, sample duration, response requirement, key peck, domestichens

Researchers using delayed matching-to-sample (DMTS) procedures have examinedthe effects of a wide range of variables on re-membering (McCarthy & White, 1987). Onevariable shown to influence DMTS perfor-mance in studies with pigeons is the responserequirement arranged on the key on whichthe sample stimulus is presented (i.e., thesample key). In general, pigeons' accuracy inmatching increases when a response is re-quired on the sample key relative to condi-tions in which no response requirement isprogrammed, and increases further as thenumber of responses required on the samplekey is increased (e.g., Lydersen, Perkins, &Chairez, 1977; Maki, Gillund, Hauge, & Sid-ers, 1977; Maki, Moe, & Bierley, 1977; Rob-erts, 1972; Roberts & Kramer, 1982; White,1985).

Increasing the duration of sample-stimuluspresentations also increases accuracy underDMTS procedures (e.g., Farthing, Wagner,Gilmour, & Waxman, 1977; Grant, 1976; Gut-tenberger & Wasserman, 1985; Leith & Maki,1975; Nelson & Wasserman, 1978; Roberts &Grant, 1974; Shimp & Moffitt, 1977). Dura-tion of sample-stimulus presentations and

The reported data were collected as part of a master'sthesis completed by C. Mackenzie at the University ofWaikato. Manuscript preparation was supported by theAcademic Research Visitor's Fund of the University ofWaikato.

Please address correspondence to Mary Foster, De-partment of Psychology, Private Bag, University of Wai-kato, Hamilton, New Zealand.

sample-key response requirements are, how-ever, characteristically confounded in studiesostensibly concerned with either variablealone. In studies of response requirements,the duration of the sample stimulus charac-teristically is determined by the time requiredto complete the programmed ratio. There-fore, as Maki, Gillund, Hauge, and Siders(1977) pointed out, increasing the responserequirement on the sample key may improveperformance wholly or in part by increasingthe duration of sample-stimulus presentation.A similar point was made by Spetch and Treit(1986), who concluded that increasing sam-ple-key response requirements increases ac-curacy primarily as a result of increased ex-posure to the sample rather than as a resultof the effort required to respond.The primary purpose of the present study

was to examine the contributions of sampleduration and sample-key response require-ments to DMTS performance. Domesticchickens were used as subjects. This specieshas been reported to respond well and simi-larly to pigeons under matching-to-sampleprocedures with no delay (DeMello, Foster, &Temple, 1992, 1993). Moreover, unpublisheddata indicate that the two species also re-spond similarly under DMTS procedures(e.g., Carroll, 1989;Jones, 1988; Odey, 1991).In the present study, hens were initiallytrained to peck the sample stimulus until atone was sounded and then to stop peckingfor the remaining time the stimulus was on.

19

1995, 64, 19-31 NUMBER I (JULY)

T. M. FOSTER et al.

One series of conditions involved no peckingon the sample, with the duration of the sam-ple stimulus varied. If the hen respondedduring the sample, then that trial was abort-ed. The aim of this series was to determinethe effects of sample duration alone. In theother conditions, requirements of 1, 3, 7, and10 responses on the sample key were com-bined with stimulus durations of 2 s, 5 s, and10 s. Trials were aborted unless exactly therequired number of responses were complet-ed while the sample stimulus was present.Care was taken to see that the required num-ber of responses could be completed withinthe stimulus duration.

METHODSubjects

Six domestic hens, numbered 51 to 56,served as subjects. Hens 52 to 56 had previ-ously served in DMTS experiments with thesame stimuli used in the present study. Hen51 was experimentally naive at the start of theexperiment. All hens were maintained at ap-proximately 80% of their free-feeding bodyweights by providing supplementary feedingof laying mash in their home cages followingexperimental sessions. Water and oyster-shellgrit were freely available in their home cages.All hens had red and fleshy combs and oc-casionally laid eggs during the course of thestudy, suggesting that they were healthy.

ApparatusThe experimental chamber (670 mm wide

by 414 mm deep by 570 mm high) was situ-ated away from the solid-state programmingequipment that controlled and recorded allexperimental events. There were three re-sponse keys in the chamber, each 34 mm indiameter, arranged horizontally 48 mm apartand 380 mm from the grid floor. The keyscould be illuminated by either red or green1-W lights. Key pecks required a minimumforce of 0.2 N, caused a 0.025-s feedbackbeep, and darkened the key for 0.025 s. Di-rectly under the center key was an aperturethat gave access to a food magazine. The mag-azine aperture could be illuminated alone orcould be illuminated with the magazineraised to provide access to wheat. A 0.5-Wwhite light (houselight), centrally located inthe ceiling of the chamber, provided ambient

illumination. When desired, a tone could besounded inside the chamber via a speakermounted on the rear wall.

ProcedureHens 52 through 56, which had DMTS his-

tories, initially were trained to stop peckingthe center key when a tone sounded for 0.5s. Pecking in Hen 51 initially was shaped bysuccessive approximations; she was then ex-posed to the same training procedure as theother birds. During training, the center (sam-ple) key was illuminated red or green for 5 swith the houselight on. When a bird peckedthis key five times within 5 s, a tone soundedfor 0.5 s, and the center keylight went out atthe end of the 5-s sample presentation. Aftera delay (0.01 s during this initial training),the two side keys were illuminated, one redand one green. The side on which each coloroccurred was determined randomly. A singlepeck to the side key that matched the previ-ous sample stimulus, termed a correct re-sponse, extinguished the keylights and pro-duced either 3-s access to the raised andilluminated magazine or 3-s illumination ofthe magazine light. A single peck to the sidekey that did not match the previous compar-ison stimulus, termed an incorrect response,resulted in a 3-s darkening of all chamberlights. If the bird pecked the center key moreor less than five times during the 5-s samplepresentation, the trial ended after 5 s with a3-s darkening of all chamber lights (black-out). Such trials were termed aborted trials.Regardless of the bird's performance, therewas no intertrial interval; the next trial start-ed as soon as food delivery, illumination ofthe magazine aperture, or darkening of thechamber ended.Throughout the experiment, the availabil-

ity of food following correct red and greenresponses was controlled by a variable-inter-val schedule with an average interval of 15 s.A controlled reinforcement procedure wasused. Thus, when food became available forone alternative (e.g., a correct response tored), the schedule stopped until that rein-forcer was collected. Correct responses thatdid not produce food were followed by 3-sillumination of the magazine. The order ofthe colors presented on the sample key waspseudorandomly determined, with the re-striction that each color could occur on no

DMTS IN HENS

Table 1

Sequence of experimental conditions for all subjects. Thenumber of sample-key responses required and the sam-ple duration under each condition are noted, as is thenumber of sessions of exposure.

Sample-key Sample Number ofCondition responses duration (s) sessions

1 3 5 202 1 5 33

Training - 5 153 7 5 414 3 2 405 1 2 35

Training 10 146 3 10 337 1 10 53

Training 10 88 7 10 219 10 10 3110 3 5 25

Training 2 1811 0 2 2712 0 5 3113 0 10 34

more than three consecutive trials. The prob-ability of a red or a green sample stimulusoccurring on any trial was as close as possibleto .5. In all conditions of the experiment, a

single daily session was conducted for eachsubject 5 days per week, at about the same

time each day.The initial training conditions were in ef-

fect for 62 sessions. The number of sample-key pecks required was then reduced tothree, with the sample duration unchangedat 5 s. After 28 sessions, the interval from theoffset of the sample stimulus to the onset ofthe comparison stimuli was increased in 0.25-s steps over the next 42 sessions to 1.25 s. Thedelay was then increased to 1.5 s for the restof the experiment. At this point, the experi-ment proper began. Each daily session duringthe experiment proper terminated after 30 or

31 deliveries of wheat for Hens 52 through56. For Hen 51, sessions were terminated af-ter 40 min if 30 wheat deliveries had not beenobtained.Performance of each hen was examined

across 13 conditions, in which sample-key re-

sponse requirements and sample durationswere varied. Response requirements of 0, 1,3, 7, and 10 and stimulus durations of 2 s, 5s, and 10 s were used, although not all pos-sible combinations were examined. Table 1

shows the order of experimental conditionsand the response requirements and sampledurations during each condition. Additionaltraining, gradually increasing or decreasingthe number of responses required to the sam-ple key, took place between some conditions.When this training occurred and the numberof days it was in effect are shown in Table 1.Each condition continued until two criteria

were met by all subjects. One required thatthere be no visually evident trend in accuracyacross sessions. The other required that themedian of the last five sessions' proportionscorrect for each subject be within .05 of themedian of the preceding five sessions' dataon five, not necessarily consecutive, occa-sions. Table 1 shows the number of sessionsrequired to meet these criteria under eachcondition.The numbers of correct and incorrect left

and right side-key responses following bothred- and green-sample trials were recordedeach session, together with the number ofwheat deliveries following red-sample trialsand the total number ofwheat deliveries. Thenumber of trials aborted due to insufficientresponses and the total number of trialsaborted also were recorded.

Data AnalysisMost data from DMTS studies have been

analyzed in terms of the percentage of cor-rect responding. This measure potentiallyconfounds the discriminability of the stimuliwith any response bias toward a particularstimulus. To separate these sources of con-trol, Davison and Tustin (1978) proposed anextension of the generalized matching equa-tion, which was used to analyze the presentdata. Describing DMTS data in the mannerproposed by Davison and Tustin is advanta-geous in that it allows for calculation of anunbiased estimate of the discriminability ofthe two stimuli, log d, and an estimate of anyinherent bias the organism may have for aparticular stimulus, log c. To calculate thesemeasures, events in the DMTS procedure arecategorized as in Figure 1. The numbers ofresponses obtained in each cell are designat-ed BSBX,B, By, and B2, and the numbers of re-inforcers are designated Rd,, Rx, R,, and R].Davison and Tustin suggested that the behav-ior following presentation of one samplestimulus, SI, may be described as

21

T. M. FOSTER et al.

Response 1 Response 2

Stimulus 1 W X

Stimulus 2 Y Z

Fig. 1. Possible events under a DMTS procedure asused in the present experiment. Stimulus 1 and Stimulus2 designate the two comparison stimuli presented on agiven trial (red, green), Response 1 and Response 2 des-ignate the two responses possible on a given trial (peckred, peck green), and the letters W, X, Y, and Z denotethe number of events (responses and reinforcers) occur-ring in each cell.

log(B/B,) = a log(k/R,)+ log c + log d, (1)

and that behavior following the presentationof the other stimulus, S2, may be described as

log(B /B,) = a log(&/R,)+ log c- log d. (2)

The parameter a measures the sensitivity ofchanges in the behavior ratio to changes inthe reinforcement ratio, log c gives a measureof inherent bias, and log d provides a mea-sure of the discriminability of the samplestimuli. A point estimate of log d can be ob-tained by subtracting Equation 2 from Equa-tion 1 to get

log d = 0.5 log[(B,.B,)/(BX.-BY)]. (3)An estimate of overall bias, which combinesthe bias due to any reinforcement asymmetryand inherent bias and which is independentof discriminability, can be obtained by addingEquations 1 and 2 to get

0.5 log [ (Bw. By) / (B,* B)]= a log(R,/R,) + log c. (4)

Controlling reinforcement frequencies, sothat the ratio of R& to Rg is 1.00, provides abehavioral measure of inherent bias, log c.When log c equals zero, percentage correctand log d will covary directly, but not neces-sarily linearly. The relationship is not neces-sarily linear, because percentage correct hasan upper limit (100%) whereas log d doesnot.

Further discussion of the use of the gen-eralized matching equation in analyzingDMTS performance is provided by Davisonand McCarthy (1988) and McCarthy andWhite (1987). Because some readers will bemore familiar with DMTS data expressed aspercentage control, a summary of the presentfindings expressed in this way is provided inAppendix A, although these data are not dis-cussed herein. The data from which log d andpercentage correct measures were derivedare provided in Appendixes B and C.

RESULTSFor each subject, data summed over the

last five sessions of each condition were usedin the following analyses. Figure 2 shows thepoint estimates of log d, calculated usingEquation 3, for each condition for each sub-ject plotted as a function of the response re-quirement on the sample key. Only complet-ed trials were used in this calculation. Thelines join data from conditions with the samesample-stimulus durations. The figure showsthat log d generally increased as the responserequirement increased. This relation held forall of the 18 functions shown. Log d usuallywas higher at the longer sample-stimulus du-rations, but the functions for different sampledurations overlap considerably.

Figure 3 shows the same data plotted as afunction of the sample-stimulus duration.The lines join data from conditions with thesame response requirements. This figureshows that, in general, as the sample-stimulusduration increased at any one response re-quirement, so did log d. In addition, for mostsubjects, the functions at each response re-quirement show reasonably clear separation,being higher with larger sample-key responserequirements.

In summary, visual analysis of the data inFigures 2 and 3 shows that DMTS perfor-mance became more accurate with increases

22

511.75

1.50

1.25

1.00

0.75

0.50

0.25

0.00

DMTS IN HENS

52

0-a.-

Io.&^l -

54 55

....0...

Y' "0,

l I I I I I I I l l

0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10

Response RequirementFig. 2. Log d values for 6 individual hens, calculated as described in the text, as a function of the response

requirement programmed on the sample key under a DMTS procedure. Separate data paths are presented for eachof three sample durations. One condition (3-s sample duration, FR 3 response requirement) was replicated; the datapoint for the replication is not connected to other points.

in both the sample-stimulus duration and theresponse requirement. To quantify these re-

lations, regression lines were fitted by themethod of least squares to these data. Slopesof these lines, shown in Table 2, provide an

indication of the relative effects on accuracyof increasing sample duration at particularresponse requirements and of increasingfixed-ratio (FR) size at particular sample du-rations.

Slopes differed considerably across hensunder a given condition. All slopes were pos-itive, indicating that accuracy (log d) alwaysincreased with increases in sample durationand FR size. Slopes indexing the effects ofsample duration did not change in gradedfashion as a function ofFR size. In 5 of 6 hens(55 was the exception), the effects of sampleduration on accuracy (i.e., slopes) were great-est when the FR size was seven. There were

23

53

0

1.75 -

1.50 -

1.25 -

1.00 -

0.75 -

0.50 -

0.25 -

0.00 -

56

T. M. FOSTER et al.

51 52

0.9

.- .-_0 - _~

1-l I I I I I I

55

.0

°O''., ;o>'' JU- .

. I I I 1

56

00.

,...0_ _UW -

_ ,_ _

0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10

Sample Duration (Seconds)Fig. 3. Log d values for 6 individual hens, calculated as described in the text, as a function of the stimulus duration

on the sample key under a DMTS procedure. Separate data paths are presented for each of three sample durations.One condition (3-s sample duration, FR 3 response requirement) was replicated; the data point for the replicationis not connected to other points.

only two data points for each bird at FR 7,however, so slopes at this value should be in-terpreted with caution.

In 5 of 6 hens (not 53), slopes indexingthe effects of FR size on accuracy were great-est when the sample duration was 2 s. In 4 ofthese birds (not 52), slopes decreased pro-gressively as the sample duration increased.These data indicate that the effects of FR size

and sample duration were not completely in-dependent but interacted somewhat.A controlled reinforcement procedure was

used in the present study; as a result, therewere almost exactly equal numbers of rein-forcements for correct responses to red andgreen stimuli in each condition. Point esti-mates of inherent bias, log c, were calculatedusing Equation 4 and taking the ratios of the

ResponseRequirement0 0* I0 3. 70 10

531.75 -

1.50 -

1.25 -

1.00-

0.75 -

0.50 -

0.25 -

0.00 -

S

.9

Q ..: ...

II I I I I I -r 1

541.75 -

1.50 -

1.25 -

1.00 -

0.75 -

0.50 -

0.25 -

0.00 -

.00

U'0

40''....'''..'''.011"_._I

,I I I I I I I

24

I II. I . I . I I .I I I II. I . II

DMTS IN HENS

Table 2Slopes of regression lines fitted to data plotted as the effects on accuracy (log d) of increasingsample duration at particular FR values (upper section) and of increasing FR size at particularsample durations (lower section).

Hen

51 52 53 54 55 56 M

FR size 0 0.19 0.31 0.42 0.58 1.34 0.24 0.511 0.19 0.19 0.21 0.24 0.86 0.16 0.313 0.20 0.12 0.32 0.06 0.90 0.14 0.297 0.83 0.89 0.86 0.90 0.55 0.37 0.73

Sample duration 2 s 1.23 0.90 0.66 1.63 1.87 1.65 1.325 s 0.78 0.49 0.51 0.76 0.90 1.33 0.8010 s 0.66 0.72 0.98 0.62 0.21 0.96 0.69

numbers of reinforcements to be 1.0. The es-timates were found to be near zero for allsubjects in all conditions; they did not changeacross response requirements or sample du-rations.Throughout the experiment, trials were

aborted when there were too few or too manyresponses during the sample presentations.Figure 4 shows the percentage of trials abort-ed for each sample duration as a function ofthe response requirement for each bird. Ingeneral, about 40% to 50% of trials wereaborted by all birds under most conditions.The aborted-trials data for Hen 51 are theexception, being much higher under mostconditions. For all birds, the number ofaborted trials generally increased as the re-sponse requirement increased. Further anal-ysis of the aborted-trials data showed that thiswas a product of increases in the number oftrials aborted as a result of too few pecks. Thepercentage of trials with too few pecks didnot increase or decrease systematically as thesample duration increased. For all subjects,there were fewer trials aborted due to toomany pecks than due to too few pecks, andthe relative number of trials with too manypecks was not obviously related to either re-sponse requirement or sample duration.Comparing the data in Figure 4 to those in

Figures 2 and 3 suggests that log d generallyincreased as the number of aborted trials in-creased. To determine the quantitative extentof the relationship between aborted trials andaccuracy, a Pearson product-moment corre-lation coeffecient was calculated between logd and percentage aborted trials across all ex-perimental conditions. The resultant rwas .48(df= 76). Although this value is significant at

a < .01, variation in the percentage of abort-ed trials accounted for only 23% of the vari-ance in accuracy.

DISCUSSIONThe present study supports unpublished

findings that indicate that domestic hens per-form relatively well, and similarly to pigeons,under DMTS procedures (e.g., Carroll, 1989;Jones, 1988; Odey, 1991). A prior study re-ported that increasing the sample-key re-quirement from 1 to 5 and 10 progressivelyincreased log d values in hens performing avisual acuity task under a matching-to-sampleprocedure with no delay, although further in-creases in response requirements did notconsistently increase performance (DeMelloet al., 1993). Response requirements above10 were not examined in the present study,in which log d increased with sample-key re-sponse requirements; thus, the present find-ings are consistent with those of DeMello etal. (1993).

Prior studies with species other than chick-ens have demonstrated that DMTS accuracyimproves when either sample durations orsample-key response requirements increase(e.g., Farthing et al., 1977; Grant, 1976; Gut-tenberger & Wasserman, 1985; Leith & Maki,1975; Lydersen et al., 1977; Maki, Gillund,Hauge, & Siders, 1977; Maki, Moe, & Bierley,1977; Nelson & Wasserman, 1978; Roberts,1972; Roberts & Grant, 1974; Roberts & Kra-mer, 1982; Shimp & Moffitt, 1977; White,1985). Similar findings were reproduced withchickens in the present study. The presentstudy differed procedurally from prior inves-tigations by neither allowing the sample du-

25

T. M. FOSTER et al.

51

54 55

I .... C

0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10

Response RequirementFig. 4. Percentage of trials aborted for 6 individual hens under all experimental conditions. A trial was aborted

if a hen pecked the sample key more or less often than the indicated response requirement while the sample keywas illuminated.

ration to vary as a function of response re-

quirements nor allowing the number ofsample-key responses to vary as a function ofsample duration. Unless this is done, it is pos-sible that the effects of increasing sample du-ration actually result from increased respond-ing or, similarly, that the effects of increasingsample-key response requirements resultfrom longer sample presentations. Neither ofthese confounding effects was possible in thepresent study, which provides evidence thatsample duration and sample-key response re-

quirements directly and independently affectDMTS performance.

Because response requirements and sam-

ple durations are scaled along different di-mensions, it is impossible to compare themagnitude of their effects in a meaningfulway. It is perhaps noteworthy that sample du-ration produced any detectable effects on ac-

curacy in the present study. Under the pro-cedures used, if few responses were requiredand the sample duration was rather long, thebird may have kept from responding by turn-

52100

50

0

100

50

0

53

56

SampleDuration0 2s* 5so lOs

26

DMTS IN HENS 27

ing away from the sample stimulus, therebylessening the actual period of exposure to(sensory contact with) the stimulus. We didnot observe the hens to determine whetherthey turned away from the sample stimulusafter completing the ratio requirement but,even if such behavior did occur, it did notprevent nominal sample duration from af-fecting accuracy.

Several theoretical models of DMTS per-formance have been developed, but their ad-equacy in accounting for the present findingsis difficult to determine. For example, Rob-erts (1972) and Roberts and Grant (1976)proposed that DMTS performance is the re-sult of comparisons of the test stimuli with agradually decaying internal analogue (trace)of the sample stimulus. It has been suggestedthat trace strength is a positive function ofsample duration and a negative function oftime since sample offset (Guttenberger &Wasserman, 1985; Roberts, 1972; Spetch &Treit, 1986).Responding to the sample stimulus might

also increase trace strength. It could be rea-sonably proposed that it is not the time thata sample stimulus is presented per se that de-termines trace strength but rather the timethat an organism attends to that stimulus. Im-posing a response requirement on the samplekey requires attention to that key, and in-creasing the response requirement may in-crease the time of forced attention. Inciden-tally, it is not necessary to infer a decayingtrace to account for the effects of increasedattention on performance. In a review ofstudies in which multiple responses to a train-ing stimulus resulted in greater discrimina-tion, Honig and Urcuioli (1981) proposedthat attention to the stimulus dimension un-der investigation is enhanced through repet-itive responding by reducing competitionfrom other stimuli. Perhaps this mechanismaccounted, at least in part, for the effects ofincreasing response requirements in the pres-ent study.

Sacks, Kamil, and Mack (1972) proposedthat increasing either response requirementsor sample durations increases accuracy by in-creasing the relative costliness of incorrect re-sponding. Log d data from the present studyare consistent with this analysis, but aborted-trials data are not. Although the hens becamemore accurate on completed trials as re-

sponse requirements increased, they becameless accurate in completing the response re-quirements. Trials with larger response re-quirements require more effort; hence, a fail-ure to complete such a trial successfully oncesome responding has occurred is in a sensemore costly than erring on a trial with lowerresponse requirements. Therefore, accordingto the analysis proposed by Sacks et al., ac-curacy in completing the sample-key re-sponse requirement should have increased,not decreased, as a function of increasing thesample-key response requirement.

In general, DMTS accuracy in the presentstudy increased as the percentage of abortedtrials increased, although variation in the lat-ter variable accounted for less than 25% ofthe total variance. In general, DMTS accuracyimproves as the intertrial interval is increased(e.g., D'Amato, 1973; Grant, 1976), and it ispossible that aborted trials in the presentstudy, which did not present both sample andcomparison stimuli, were functionally equiv-alent to intertrial periods.Examining DMTS data in terms of the gen-

eralized matching law, as was done in thepresent study, allows for a finer grained anal-ysis than examining such data in terms of per-centage of correct responding (Davison &McCarthy, 1988; Davison & Tustin, 1978; Mc-Carthy & White, 1987). The conclusions sup-ported by the two kinds of analyses do notnecessarily agree, but they do so in the caseof the present data. For the present data, es-timates of inherent bias were close to zeroand did not change across conditions. There-fore, changes in log d values closely approxi-mated changes in overall percentage of cor-rect responding values and supportequivalent conclusions. Such an outcome isnot foregone, however, and describing DMTSdata in the manner proposed by Davison andTustin may well prove to be advantageous infuture studies.

REFERENCES

Carroll, I. G. (1989). Delayed matching to sample: Effects ofsample duration. Unpublished master's thesis, Univer-sity of Waikato, Hamilton, New Zealand.

D'Amato, M. R. (1973). Delayed matching and shortterm memory in monkeys. In G. H. Bower (Ed.), Thepsychology of learning and motivation: Advances in researchand theory (Vol. 7, pp. 227-269). New York: AcademicPress.

28 T. M. FOSTER et al.

Davison, M. C., & McCarthy, D. (1988). The matching law:A research review. Hillsdale, NJ: Erlbaum.

Davison, M. C., & Tustin, R. D. (1978). On the relationbetween the generalized matching law and signal-de-tection theory. Journal of the Experimental Analysis ofBe-havior, 29, 331-336.

DeMello, L. R., Foster, T. M., & Temple, W. (1992). Dis-criminative performance of the domestic hen in a vi-sual acuity task. Journal of the Experimental Analysis ofBehavior, 58, 147-157.

DeMello, L. R., Foster, T. M., & Temple, W. (1993). Theeffect of increased response requirements on discrim-inative performance of the domestic hen in a visualacuity task. Journal of the Experimental Analysis ofBehav-ior, 60, 595-609.

Farthing, G. W., Wagner, J. M., Gilmour, S., & Waxman,H. M. (1977). Short-term memory and informationprocessing in pigeons. Learning and Motivation, 8,520-532.

Grant, D. S. (1976). Effect of sample presentation timeon long-delay matching in the pigeon. Learning andMotivation, 7, 580-590.

Guttenberger, V. T., & Wasserman, E. A. (1985). Effectsof sample duration, retention interval and passage oftime in the test on pigeons' matching-to-sample per-formance. Animal Learning & Behavior, 13, 121-128.

Honig, W. K, & Urcuioli, P. J. (1981). The legacy ofGuttman and Kalish (1956): 25 years of research onstimulus generalization. Journal of the ExperimentalAnalysis of Behavior, 36, 405-445.

Jones, B. M. (1988). Response-produced stimuli in a delayedmatching to sample task. Unpublished master's thesis,University of Waikato, Hamilton, New Zealand.

Leith, C. R., & Maki, W. S., Jr. (1975). Attention shiftsduring matching-to-sample performance in pigeons.Animal Learning & Behavior, 3, 85-89.

Lydersen, T., Perkins, D., & Chairez, H. (1977). Effectsof fixed-ratio sample and choice response require-ments upon oddity matching. Journal of the Experimen-tal Analysis of Behavior, 25, 97-101.

Maki, W. S., Jr., Gillund, G., Hauge, G., & Siders, W. S.(1977). Matching to sample after extinction of ob-serving responses. Journal of Experimental Psychology:Animal Behavior Processes, 3, 285-296.

Maki, W. S.,Jr., Moe,J. C., & Bierley, C. M. (1977). Short-term memory for stimuli, responses and reinforcers.Journal of Experimental Psychology: Animal Behavior Pro-cesses, 3, 156-177.

McCarthy, D., & White, K. G. (1987). Behavioral modelsof delayed detection and their application to thestudy of memory. In M. L. Commons, J. E. Mazur, J.A. Nevin, & H. C. Rachlin (Eds.), Quantitative analysesof behavior: Vol. 5. The effect ofdelay and interveningeventson reinforcer value (pp. 29-54). Hillsdale, NJ: Erlbaum.

Nelson, K. R., & Wasserman, E. A. (1978). Temporal fac-tors influencing the pigeon's successive matching-to-sample performance: Sample duration, intertrial in-terval and retention interval.Journal of the ExperimentalAnalysis of Behavior 30, 153-162.

Odey, M. (1991). D.M.TS. zvith hens: The effects of someprocedural variations Unpublished master's thesis, Uni-versity of Waikato, Hamilton, New Zealand.

Roberts, W. A. (1972). Short-term memory in the pi-geon: Effects of repetition and spacing. Journal ofEx-perimental Psychology, 94, 74-83.

Roberts, W. A., & Grant, D. S. (1974). Short-term mem-ory in the pigeon with presentation precisely con-trolled. Learning and Motivation, 5, 393-408.

Roberts, W. A., & Grant, D. S. (1976). Studies of shortterm memory in the pigeon using the delayed match-ing-to-sample procedure. In D. L. Medin, W. A. Rob-erts, & R. T. Davis (Eds.), Processes of animal memory(pp. 79-112). Hillsdale, NJ: Erlbaum.

Roberts, W. A., & Kramer, P. J. (1982). Some observa-tions of the effects of intertrial interval and delay ondelayed matching to sample in pigeons. Journal ofEx-perimental Psychology: Animal Behavior Processes, 8, 342-353.

Sacks, R. A., Kamil, A. C., & Mack, R. (1972). The effectsof fixed-ratio requirement on matching to sample inthe pigeon. Psychonomic Science, 26, 291-293.

Shimp, C. P., & Moffitt, M. (1977). Short-term memoryin the pigeon: Delayed-pair-comparison proceduresand some results. Journal of the Experimental Analysis ofBehavior, 28, 13-25.

Spetch, M. L., & Treit, D. (1986). Does effort play a rolein the effect of response requirements on delayedmatching to sample? Journal of the ExperimentalAnalysisof Behavior, 45, 19-31.

White, K G. (1985). Characteristics of forgetting func-tions in delayed matching to sample. Journal of the Ex-perimental Analysis of Behavior, 44, 15-34.

Received August 8, 1994Final acceptance March 9, 1995

APPENDIX AMean percentage correct by each hen during the last five sessions of exposure to all experi-mental conditions.

Sample duration

2 s 5 s 10 s

Hen FR0 FRI FR3 FR0 FRI FR3 FR7 FR0 FRI FR3 FR7 FR10

51 55 75 76 52 77 67,79 84 62 81 81 93 8952 61 76 76 64 78 85, 86 82 73 81 82 98 9153 65 81 76 70 70 80, 87 84 80 85 86 94 9854 57 75 82 75 75 83,86 90 80 82 84 94 9355 64 85 88 84 79 87,92 94 96 96 97 97 9756 54 70 79 53 67 77, 87 91 62 75 83 94 93

DMTS IN HENS

APPENDIX BRaw data from which log d and percentage correct measures were derived. All data are totalsfor individual hens across the final five sessions of exposure to the listed condition.

Trials Correct responses Incorrect responsesaborted with

too few Total trials Red on Red on Red on Red onCondition FR Sample (s) responses aborted left key right key left key right key

1 3 5 9391 3 5 1661 3 5 1381 3 5 2061 3 5 1161 3 5 2352 1 5 8782 1 5 1262 1 5 662 1 5 1692 1 5 972 1 5 1463 7 5 1,1343 7 5 3253 7 5 2573 7 5 2553 7 5 1,1603 7 5 4124 3 2 2,0064 3 2 3144 3 2 3184 3 2 2304 3 2 3834 3 2 3465 1 2 5165 1 2 1645 1 2 2175 1 2 1635 1 2 2015 1 2 2966 3 10 2816 3 10 916 3 10 556 3 10 1336 3 10 966 3 10 1727 1 10 1717 1 10 1317 1 10 138

1,175249162243245288

1,047220102211148163

1,239355283263

1,174454

2,070350387264412390671185259221244317

1,036179154188190277276176169

52 2887 8297 7876 8795 8778 7462 3860 100107 7956 11085 7854 14360 4675 8580 9471 9569 6271 7870 49104 6482 9375 10076 11781 111105 63117 88115 11083 121106 8272 13451 3655 9171 8573 8683 6666 10193 6662 9980 77

3319371412113415601724157

2414561

41432223145

5542241824161012151235

281021

21116238

264

3920562973878143

119

228

2514493

202047157293222216

357

4110

29

T. M. FOSTER et al.

APPENDIX B

(Continued)

Trials Correct responses Incorrect responsesaborted with

too few Total trials Red on Red on Red on Red onCondition FR Sample (s) responses aborted left key right key left key right key

7 1 10 977 1 10 1137 1 10 428 7 10 5088 7 10 1868 7 10 1288 7 10 1658 7 10 2698 7 10 4909 10 10 6199 10 10 2409 10 10 3489 10 10 1869 10 10 2679 10 10 36810 3 5 90610 3 5 19910 3 5 11010 3 5 5910 3 5 19710 3 5 15611 0 2 011 0 2 011 0 2 011 0 2 011 0 2 011 0 2 012 0 5 012 0 5 012 0 5 012 0 5 012 0 5 012 0 5 013 0 10 013 0 10 013 0 10 013 0 10 013 0 10 013 0 10 0

12814214180928120917632352591429137818730043696127415285

302210

92045217

11277

405

4069914

16529

87 10582 8667 13079 6179 6991 8264 7789 7181 7554 4771 6262 6868 6571 7893 8957 7364 8292 7191 9089 7190 82144 6551 14473 11033 158

151 5443 145124 4570 10161 8564 119120 69123 4898 7051 11371 8596 8693 8075 88

197

1711111064914121707681751014

122194926

1114413444441532

111707

28149

53

2443403

1633331

1521056

22227

34814381129413920150226478430

55374823293

45

30

DMTS nv HENS

APPENDIX CRaw data from which log d and percentage correct mea-sures were derived. All data are totals for individual hensacross the final five sessions of exposure to the listed con-dition.

Incorrect Correctresponses responses Food

Green Green Green Green deliv- Totalon on on on eries food

Condi- left right left right on red deliv-tion key key key key trials eries

1 43 6 68 28 45 861 20 12 93 91 80 1521 21 25 111 105 81 1551 11 21 95 88 82 1541 21 11 113 74 81 1551 14 36 77 71 81 1552 24 2 72 49 61 1182 3 40 66 121 77 1552 50 31 98 94 78 1552 7 40 68 124 79 1552 24 17 103 83 78 1542 7 85 50 117 78 1543 24 0 59 44 52 1073 21 16 93 64 77 1553 18 25 98 62 78 1553 2 13 78 72 78 1543 3 2 53 54 60 1173 4 16 84 84 78 1554 32 4 86 68 61 1264 35 15 112 92 78 1544 36 17 124 99 79 1544 18 19 116 85 78 1534 17 5 99 84 77 1554 6 39 79 105 76 1545 55 3 122 61 75 1545 39 24 98 96 75 1555 28 24 90 99 79 1545 15 58 90 122 80 1555 21 6 105 88 80 1555 25 66 73 131 75 1556 17 8 55 50 52 986 6 22 73 104 77 1556 1 17 65 107 79 1546 8 23 86 92 77 1556 2 0 75 81 78 1556 5 23 85 87 77 1557 35 3 100 55 75 1557 6 23 67 121 75 1557 21 8 95 85 80 1557 6 31 80 97 80 155

APPENDIX C

(Continueda)

Incorrect Correctresponses responses Food

Green Green Green Green deliv- Total

Condi-tion

77888888999999101010101010111111111111121212121212131313131313

on onleft rightkey key

3 118 446 03 27 41 81 22 69 53 12 21 61 29 110 386 1910 204 167 414 11

137 3421 10227 6023 11892 1841 122127 3039 5712 246 56

27 15115 4060 339 57

21 1217 255 2

53 53

on on eries foodleft right on red deliv-key key trials eries

81 73 79 15549 109 76 15569 53 73 15072 84 76 15585 87 80 15569 79 80 15572 74 79 15573 79 75 15560 62 57 11285 79 76 155107 82 79 15576 90 79 15578 64 79 15571 71 75 15562 89 70 14395 108 75 15595 88 80 15591 119 80 15595 78 80 15594 81 74 154148 42 74 15367 139 75 15583 153 80 15443 168 80 154142 84 79 15464 158 80 155134 31 75 155102 96 75 155109 118 79 15552 126 78 153107 88 80 155128 63 76 15590 71 75 15567 101 75 15587 94 79 15576 90 80 155110 121 80 15588 82 76 155

===